diff --git a/.gitattributes b/.gitattributes index 4da5a4c9997202ac5828e9a5a075feda8119a927..1eaaf11267027e33958eb844a80c04de6e65b90f 100644 --- a/.gitattributes +++ b/.gitattributes @@ -74,3 +74,4 @@ qwen35-2b-fp32-int8-g32-exclude-l6-l9-l10up-l15o-home-token-major-v2-v4/tokenize qwen35-2b-multimodal-fp32-three-grid-state-alias-token-major-v2-v4/graph.json filter=lfs diff=lfs merge=lfs -text qwen35-2b-multimodal-fp32-three-grid-state-alias-token-major-v2-v4/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text qwen35-08b-fp32-8k-token-major-home-v5/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text +qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/graph.json filter=lfs diff=lfs merge=lfs -text diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/entrypoints/prefill-16.json b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/entrypoints/prefill-16.json new file mode 100644 index 0000000000000000000000000000000000000000..e1d64d84d86633612580a0baf9becc45ad7cee15 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/entrypoints/prefill-16.json @@ -0,0 +1,83074 @@ +{ + "inputs": [ + "input_ids", + "position_ids", + "attention_mask", + "states_0", + "states_1", + "states_2", + "states_3", + "states_4", + "states_5", + "states_6", + "states_7", + "states_8", + "states_9", + "states_10", + "states_11", + "states_12", + "states_13", + "states_14", + "states_15", + "states_16", + "states_17", + "states_18", + "states_19", + "states_20", + "states_21", + "states_22", + "states_23", + "states_24", + "states_25", + "states_26", + "states_27", + "states_28", + "states_29", + "states_30", + "states_31", + "states_32", + "states_33", + "states_34", + "states_35", + "states_36", + "states_37", + "states_38", + "states_39", + "states_40", + "states_41", + "states_42", + "states_43", + "states_44", + "states_45", + "states_46", + "states_47" + ], + "outputs": [ + "linear_186", + "slice_10", + "slice_16", + "slice_18", + "slice_24", + "slice_26", + "slice_32", + "cache_update", + "cache_update_1", + "slice_44", + "slice_50", + "slice_52", + "slice_58", + "slice_60", + "slice_66", + "cache_update_2", + "cache_update_3", + "slice_78", + "slice_84", + "slice_86", + "slice_92", + "slice_94", + "slice_100", + "cache_update_4", + "cache_update_5", + "slice_112", + "slice_118", + "slice_120", + "slice_126", + "slice_128", + "slice_134", + "cache_update_6", + "cache_update_7", + "slice_146", + "slice_152", + "slice_154", + "slice_160", + "slice_162", + "slice_168", + "cache_update_8", + "cache_update_9", + "slice_180", + "slice_186", + "slice_188", + "slice_194", + "slice_196", + "slice_202", + "cache_update_10", + "cache_update_11" + ], + "tensors": { + "p_model_model_language_model_embed_tokens_weight": { + "shape": [ + 248320, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 254279680, + "storageDtype": "uint32", + "parts": [ + { + "path": "weights/int8-422da3921400-000.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-001.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-002.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-003.bin", + "offset": 0, + "byteLength": 52953088 + } + ], + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_embed_tokens_weight__int8_scales", + "wordsPerRow": 256 + } + }, + "p_model_model_language_model_layers_0_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 38797312, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 38801408, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 38801472, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 38801536, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 45092992, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47190144, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47206528, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47222912, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47321216, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47321728, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_0_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 49418880, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 49422976, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 51257984, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 53092992, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_1_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 54928000, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 54932096, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 54932160, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 54932224, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 61223680, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63320832, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63337216, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63353600, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63451904, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63452416, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_1_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 65549568, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 65553664, + "byteLength": 1555200 + }, + { + "path": "weights/00008.bin", + "offset": 0, + "byteLength": 279808 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 279808, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 2114816, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_2_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 3949824, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 3953920, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 3953984, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 3954048, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 10245504, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12342656, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12359040, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12375424, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12473728, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12474240, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_2_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 14571392, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 14575488, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 16410496, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 18245504, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_3_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 20080512, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 20084608, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 24278912, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 24279936, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 24804224, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 24805248, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 25329536, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_3_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 27426688, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 27430784, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 29265792, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 31100800, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_4_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 32935808, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 32939904, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 32939968, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 32940032, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 39231488, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41328640, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41345024, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41361408, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41459712, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41460224, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_4_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 43557376, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 43561472, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 45396480, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 47231488, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_5_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 49066496, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 49070592, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 49070656, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 49070720, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 55362176, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57459328, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57475712, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57492096, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57590400, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57590912, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_5_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 59688064, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 59692160, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 61527168, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 63362176, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_6_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 65197184, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 65201280, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 65201344, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 65201408, + "byteLength": 1907456 + }, + { + "path": "weights/00009.bin", + "offset": 0, + "byteLength": 4384000 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 4384000, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6481152, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6497536, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6513920, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6612224, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6612736, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_6_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 8709888, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 8713984, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 10548992, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 12384000, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_7_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 14219008, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 14223104, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 18417408, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 18418432, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 18942720, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 18943744, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 19468032, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_7_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 21565184, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 21569280, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 23404288, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 25239296, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_8_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 27074304, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 27078400, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 27078464, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 27078528, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 33369984, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35467136, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35483520, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35499904, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35598208, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35598720, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_8_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 37695872, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 37699968, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 39534976, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 41369984, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_9_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 43204992, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 43209088, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 43209152, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 43209216, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 49500672, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51597824, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51614208, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51630592, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51728896, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51729408, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_9_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 53826560, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 53830656, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 55665664, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 57500672, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_10_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 59335680, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 59339776, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 59339840, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 59339904, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 65631360, + "byteLength": 1477504 + }, + { + "path": "weights/00010.bin", + "offset": 0, + "byteLength": 619648 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 619648, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 636032, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 652416, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 750720, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 751232, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_10_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 2848384, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 2852480, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 4687488, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 6522496, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_11_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 8357504, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 8361600, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 12555904, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 12556928, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 13081216, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 13082240, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 13606528, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_11_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 15703680, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 15707776, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 17542784, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 19377792, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_12_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 21212800, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 21216896, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 21216960, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 21217024, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 27508480, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29605632, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29622016, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29638400, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29736704, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29737216, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_12_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 31834368, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 31838464, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 33673472, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 35508480, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_13_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 37343488, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 37347584, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 37347648, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 37347712, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 43639168, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45736320, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45752704, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45769088, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45867392, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45867904, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_13_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 47965056, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 47969152, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 49804160, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 51639168, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_14_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 53474176, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 53478272, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 53478336, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 53478400, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 59769856, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61867008, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61883392, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61899776, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61998080, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61998592, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_14_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 64095744, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 64099840, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 65934848, + "byteLength": 1174016 + }, + { + "path": "weights/00011.bin", + "offset": 0, + "byteLength": 660992 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 660992, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_15_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 2496000, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 2500096, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 6694400, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 6695424, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 7219712, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 7220736, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 7745024, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_15_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 9842176, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 9846272, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 11681280, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 13516288, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_16_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 15351296, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 15355392, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 15355456, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 15355520, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 21646976, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23744128, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23760512, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23776896, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23875200, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23875712, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_16_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 25972864, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 25976960, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 27811968, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 29646976, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_17_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 31481984, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 31486080, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 31486144, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 31486208, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 37777664, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 39874816, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 39891200, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 39907584, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 40005888, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 40006400, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_17_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 42103552, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 42107648, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 43942656, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 45777664, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_18_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 47612672, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 47616768, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 47616832, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 47616896, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 53908352, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56005504, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56021888, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56038272, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56136576, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56137088, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_18_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 58234240, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 58238336, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 60073344, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 61908352, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_19_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 63743360, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 63747456, + "byteLength": 3361408 + }, + { + "path": "weights/00012.bin", + "offset": 0, + "byteLength": 832896 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 832896, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 833920, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 1358208, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 1359232, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 1883520, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_19_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 3980672, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 3984768, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 5819776, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 7654784, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_20_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 9489792, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 9493888, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 9493952, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 9494016, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 15785472, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 17882624, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 17899008, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 17915392, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 18013696, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 18014208, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_20_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 20111360, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 20115456, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 21950464, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 23785472, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_21_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 25620480, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 25624576, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 25624640, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 25624704, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 31916160, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34013312, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34029696, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34046080, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34144384, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34144896, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_21_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 36242048, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 36246144, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 38081152, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 39916160, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_22_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 41751168, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 41755264, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 41755328, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 41755392, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 48046848, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50144000, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50160384, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50176768, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50275072, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50275584, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_22_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 52372736, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 52376832, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 54211840, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 56046848, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_23_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 57881856, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 57885952, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 62080256, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 62081280, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 62605568, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 62606592, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 63130880, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_23_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 65228032, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 65232128, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 67067136, + "byteLength": 41728 + }, + { + "path": "weights/00013.bin", + "offset": 0, + "byteLength": 1793280 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 1793280, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_norm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 3628288, + "byteLength": 4096 + } + ] + }, + "b_model_model_language_model_rotary_emb_inv_freq": { + "shape": [ + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 128, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 3632384, + "byteLength": 128 + } + ] + }, + "input_ids": { + "shape": [ + 1, + 16 + ], + "dtype": "int32", + "role": "input", + "byteLength": 64, + "indexBounds": [ + 0, + 248320 + ] + }, + "position_ids": { + "shape": [ + 1, + 16 + ], + "dtype": "int32", + "role": "input", + "byteLength": 64, + "indexBounds": [ + 0, + 4096 + ] + }, + "attention_mask": { + "shape": [ + 1, + 1, + 16, + 4096 + ], + "dtype": "float32", + "role": "input", + "byteLength": 262144 + }, + "states_0": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_1": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_2": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_3": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_4": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_5": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_6": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_7": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_8": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_9": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_10": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_11": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_12": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_13": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_14": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_15": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_16": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_17": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_18": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_19": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_20": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_21": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_22": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_23": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_24": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_25": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_26": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_27": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_28": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_29": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_30": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_31": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_32": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_33": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_34": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_35": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_36": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_37": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_38": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_39": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_40": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_41": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_42": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_43": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_44": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_45": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_46": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_47": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "embedding": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "unsqueeze": { + "shape": [ + 1, + 1, + 16 + ], + "dtype": "int32", + "role": "value", + "byteLength": 64 + }, + "expand": { + "shape": [ + 4, + 1, + 16 + ], + "dtype": "int32", + "role": "value", + "byteLength": 256 + }, + "slice_1": { + "shape": [ + 3, + 1, + 16 + ], + "dtype": "int32", + "role": "value", + "byteLength": 192 + }, + "unsqueeze_1": { + "shape": [ + 1, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "unsqueeze_2": { + "shape": [ + 1, + 1, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "unsqueeze_3": { + "shape": [ + 1, + 1, + 32, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "expand_1": { + "shape": [ + 3, + 1, + 32, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 384 + }, + "unsqueeze_4": { + "shape": [ + 3, + 1, + 1, + 16 + ], + "dtype": "int32", + "role": "value", + "byteLength": 192 + }, + "_to_copy": { + "shape": [ + 3, + 1, + 1, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 192 + }, + "matmul": { + "shape": [ + 3, + 1, + 32, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "transpose": { + "shape": [ + 3, + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "select": { + "shape": [ + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "select_1": { + "shape": [ + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_2": { + "shape": [ + 1, + 16, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 704 + }, + "slice_3": { + "shape": [ + 1, + 16, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 704 + }, + "copy": { + "shape": [ + 1, + 16, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 704 + }, + "transpose_1": { + "shape": [ + 3, + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "select_2": { + "shape": [ + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_scatter": { + "shape": [ + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "select_scatter": { + "shape": [ + 3, + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "transpose_2": { + "shape": [ + 3, + 1, + 32, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "transpose_7": { + "shape": [ + 3, + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "select_7": { + "shape": [ + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_7": { + "shape": [ + 1, + 16, + 10 + ], + "dtype": "float32", + "role": "value", + "byteLength": 640 + }, + "transpose_8": { + "shape": [ + 3, + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "select_8": { + "shape": [ + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_8": { + "shape": [ + 1, + 16, + 10 + ], + "dtype": "float32", + "role": "value", + "byteLength": 640 + }, + "copy_1": { + "shape": [ + 1, + 16, + 10 + ], + "dtype": "float32", + "role": "value", + "byteLength": 640 + }, + "transpose_9": { + "shape": [ + 3, + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "select_9": { + "shape": [ + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_scatter_1": { + "shape": [ + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "select_scatter_1": { + "shape": [ + 3, + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "transpose_10": { + "shape": [ + 3, + 1, + 32, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "transpose_12": { + "shape": [ + 3, + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "select_11": { + "shape": [ + 1, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "cat": { + "shape": [ + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cos": { + "shape": [ + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul": { + "shape": [ + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "sin": { + "shape": [ + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_1": { + "shape": [ + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "pow_1": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_2": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_1": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_3": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_13": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_1": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_2": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_3": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_1": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_10": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_5": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_11": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_12": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_14": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_13": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_14": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_15": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_1": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_2": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_3": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_2": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_4": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_5": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_1": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_3": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_1": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_6": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_7": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_2": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_4": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_2": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_8": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_16": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_17": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_15": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_4": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_5": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_2": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_1": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_5": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_3": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_9": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_10": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_1": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_11": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_6": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_4": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_6": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_3": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_2": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_7": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_4": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_12": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_8": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_13": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_1": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_5": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_2": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_6": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_14": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_7": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_9": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_4": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_3": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_10": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_5": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_15": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_11": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_16": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_2": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_8": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_16": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_9": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_7": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_10": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_11": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_2": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_18": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_1": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_6": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_1": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_19": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_3": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_20": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_17": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_21": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_22": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_23": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_8": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_9": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_10": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_1": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_1": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_1": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_12": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_1": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_17": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_18": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_3": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_13": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_6": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_19": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_20": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_4": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_14": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_7": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_21": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_1": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_24": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_25": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_18": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_1": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_11": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_12": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_5": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_4": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_15": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_8": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_22": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_23": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_4": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_24": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_13": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_12": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_16": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_6": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_5": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_17": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_9": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_25": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_18": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_26": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_3": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_13": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_5": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_14": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_27": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_15": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_19": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_7": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_6": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_20": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_10": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_28": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_21": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_29": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_4": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_16": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_19": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_17": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_14": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_18": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_19": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_3": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_26": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_2": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_7": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_2": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_27": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_6": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_28": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_20": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_29": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_30": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_31": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_15": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_16": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_17": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_2": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_2": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_2": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_22": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_2": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_30": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_31": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_5": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_23": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_11": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_32": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_33": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_6": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_24": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_12": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_34": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_2": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_32": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_33": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_21": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_2": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_18": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_19": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_8": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_7": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_25": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_13": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_35": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_36": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_7": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_37": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_20": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_20": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_26": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_9": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_8": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_27": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_14": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_38": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_28": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_39": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_5": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_21": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_8": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_22": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_40": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_23": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_29": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_10": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_9": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_30": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_15": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_41": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_31": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_42": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_6": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_24": { + "shape": [ + 1, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "view_21": { + "shape": [ + 1, + 16, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "slice_34": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_35": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_3": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "_unsafe_view": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_22": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_11": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_10": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "add_32": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "rsqrt_16": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "mul_43": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_33": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_44": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "type_as_7": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_22": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_25": { + "shape": [ + 1, + 16, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_23": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_12": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_11": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "add_34": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "rsqrt_17": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "mul_45": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_35": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_46": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_8": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_23": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_26": { + "shape": [ + 1, + 16, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_24": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_24": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "unsqueeze_8": { + "shape": [ + 1, + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "unsqueeze_9": { + "shape": [ + 1, + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_36": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_37": { + "shape": [ + 1, + 8, + 16, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_38": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_39": { + "shape": [ + 1, + 2, + 16, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "mul_47": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_40": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_41": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "neg_3": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "cat_4": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_48": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_36": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_49": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_42": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_43": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "neg_4": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_5": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_50": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_37": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "cat_6": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "cat_7": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cache_update": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_1": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_10": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_2": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_4": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_1": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_11": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_3": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_5": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_2": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_25": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_1": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "mul_51": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "add_38": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "softmax": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "matmul_2": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_26": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_6": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_25": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_3": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_52": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_27": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_39": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_13": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_12": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_40": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_18": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_53": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_41": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_54": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_9": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_28": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_9": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_29": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_55": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_30": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_42": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_14": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_13": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_43": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_19": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_56": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_44": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_57": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_10": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_31": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_27": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_32": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_26": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_33": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_34": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_8": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_44": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_3": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_12": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_3": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_45": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_10": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_46": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_28": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_47": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_48": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_49": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_27": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_28": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_29": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_4": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_3": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_5": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_45": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_3": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_58": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_59": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_7": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_46": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_20": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_60": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_61": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_8": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_47": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_21": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_62": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_3": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_50": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_51": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_29": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_7": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_30": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_31": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_15": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_14": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_48": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_22": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_63": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_64": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_11": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_65": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_32": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_35": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_49": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_16": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_15": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_50": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_23": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_66": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_51": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_67": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_11": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_36": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_12": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_37": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_68": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_38": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_52": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_17": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_16": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_53": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_24": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_69": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_54": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_70": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_12": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_39": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_30": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_40": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_33": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_41": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_42": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_9": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_52": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_4": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_13": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_4": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_53": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_13": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_54": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_31": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_55": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_56": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_57": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_34": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_35": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_36": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_5": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_4": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_6": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_55": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_4": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_71": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_72": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_9": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_56": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_25": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_73": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_74": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_10": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_57": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_26": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_75": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_4": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_58": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_59": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_32": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_8": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_37": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_38": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_18": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_17": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_58": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_27": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_76": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_77": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_14": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_78": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_39": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_43": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_59": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_19": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_18": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_60": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_28": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_79": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_61": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_80": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_13": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_44": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_15": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_45": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_81": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_46": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_62": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_20": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_19": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_63": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_29": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_82": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_64": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_83": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_14": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_47": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_33": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_48": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_40": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_49": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_50": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_10": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_60": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_5": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_14": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_5": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_61": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_16": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_62": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_34": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_63": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_64": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_65": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_41": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_42": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_43": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_6": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_5": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_7": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_65": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_5": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_84": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_85": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_11": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_66": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_30": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_86": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_87": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_12": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_67": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_31": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_88": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_5": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_66": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_67": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_35": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_9": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_44": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_45": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_21": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_20": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_68": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_32": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_89": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_90": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_17": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_91": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_46": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_51": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_69": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_22": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_21": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_70": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_33": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_92": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_71": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_93": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_15": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_52": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_18": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_53": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_94": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_54": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_72": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_23": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_22": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_73": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_34": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_95": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_74": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_96": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_16": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_55": { + "shape": [ + 1, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "view_47": { + "shape": [ + 1, + 16, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "slice_68": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_69": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_10": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "_unsafe_view_3": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_48": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_24": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_23": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "add_75": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "rsqrt_35": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "mul_97": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_76": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_98": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "type_as_17": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_36": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_56": { + "shape": [ + 1, + 16, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_49": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_25": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_24": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "add_77": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "rsqrt_36": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "mul_99": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_78": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_100": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_18": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_37": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_57": { + "shape": [ + 1, + 16, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_50": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_38": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "unsqueeze_15": { + "shape": [ + 1, + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "unsqueeze_16": { + "shape": [ + 1, + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_70": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_71": { + "shape": [ + 1, + 8, + 16, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_72": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_73": { + "shape": [ + 1, + 2, + 16, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "mul_101": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_74": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_75": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "neg_8": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "cat_11": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_102": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_79": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_103": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_76": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_77": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "neg_9": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_12": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_104": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_80": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "cat_13": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "cat_14": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cache_update_2": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_3": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_17": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_4": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_11": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_4": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_18": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_5": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_12": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_5": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_39": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_3": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "mul_105": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "add_81": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "softmax_1": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "matmul_4": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_40": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_13": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_51": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_7": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_106": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_58": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_82": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_26": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_25": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_83": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_37": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_107": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_84": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_108": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_19": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_59": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_19": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_60": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_109": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_61": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_85": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_27": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_26": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_86": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_38": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_110": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_87": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_111": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_20": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_62": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_41": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_63": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_52": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_64": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_65": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_15": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_78": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_6": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_19": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_6": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_79": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_20": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_80": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_42": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_81": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_82": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_83": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_53": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_54": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_55": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_8": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_6": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_10": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_88": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_6": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_112": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_113": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_13": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_89": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_39": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_114": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_115": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_14": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_90": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_40": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_116": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_6": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_84": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_85": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_43": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_14": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_56": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_57": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_28": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_27": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_91": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_41": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_117": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_118": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_21": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_119": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_58": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_66": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_92": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_29": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_28": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_93": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_42": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_120": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_94": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_121": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_21": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_67": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_22": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_68": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_122": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_69": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_95": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_30": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_29": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_96": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_43": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_123": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_97": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_124": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_22": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_70": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_44": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_71": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_59": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_72": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_73": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_16": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_86": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_7": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_20": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_7": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_87": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_23": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_88": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_45": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_89": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_90": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_91": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_60": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_61": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_62": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_9": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_7": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_11": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_98": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_7": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_125": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_126": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_15": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_99": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_44": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_127": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_128": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_16": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_100": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_45": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_129": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_7": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_92": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_93": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_46": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_15": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_63": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_64": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_31": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_30": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_101": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_46": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_130": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_131": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_24": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_132": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_65": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_74": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_102": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_32": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_31": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_103": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_47": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_133": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_104": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_134": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_23": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_75": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_25": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_76": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_135": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_77": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_105": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_33": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_32": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_106": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_48": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_136": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_107": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_137": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_24": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_78": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_47": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_79": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_66": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_80": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_81": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_17": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_94": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_8": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_21": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_8": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_95": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_26": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_96": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_48": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_97": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_98": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_99": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_67": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_68": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_69": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_10": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_8": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_12": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_108": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_8": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_138": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_139": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_17": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_109": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_49": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_140": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_141": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_18": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_110": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_50": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_142": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_8": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_100": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_101": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_49": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_16": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_70": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_71": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_34": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_33": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_111": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_51": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_143": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_144": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_27": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_145": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_72": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_82": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_112": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_35": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_34": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_113": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_52": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_146": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_114": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_147": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_25": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_83": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_28": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_84": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_148": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_85": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_115": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_36": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_35": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_116": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_53": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_149": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_117": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_150": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_26": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_86": { + "shape": [ + 1, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "view_73": { + "shape": [ + 1, + 16, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "slice_102": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_103": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_17": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "_unsafe_view_6": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_74": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_37": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_36": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "add_118": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "rsqrt_54": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "mul_151": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_119": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_152": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "type_as_27": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_50": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_87": { + "shape": [ + 1, + 16, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_75": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_38": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_37": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "add_120": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "rsqrt_55": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "mul_153": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_121": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_154": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_28": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_51": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_88": { + "shape": [ + 1, + 16, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_76": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_52": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "unsqueeze_22": { + "shape": [ + 1, + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "unsqueeze_23": { + "shape": [ + 1, + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_104": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_105": { + "shape": [ + 1, + 8, + 16, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_106": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_107": { + "shape": [ + 1, + 2, + 16, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "mul_155": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_108": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_109": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "neg_13": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "cat_18": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_156": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_122": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_157": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_110": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_111": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "neg_14": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_19": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_158": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_123": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "cat_20": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "cat_21": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cache_update_4": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_5": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_24": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_6": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_18": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_7": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_25": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_7": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_19": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_8": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_53": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_5": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "mul_159": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "add_124": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "softmax_2": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "matmul_6": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_54": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_20": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_77": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_11": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_160": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_89": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_125": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_39": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_38": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_126": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_56": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_161": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_127": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_162": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_29": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_90": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_29": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_91": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_163": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_92": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_128": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_40": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_39": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_129": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_57": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_164": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_130": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_165": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_30": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_93": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_55": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_94": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_78": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_95": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_96": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_22": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_112": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_9": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_26": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_9": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_113": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_30": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_114": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_56": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_115": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_116": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_117": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_79": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_80": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_81": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_12": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_9": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_15": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_131": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_9": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_166": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_167": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_19": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_132": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_58": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_168": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_169": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_20": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_133": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_59": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_170": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_9": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_118": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_119": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_57": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_21": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_82": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_83": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_41": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_40": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_134": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_60": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_171": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_172": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_31": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_173": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_84": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_97": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_135": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_42": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_41": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_136": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_61": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_174": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_137": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_175": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_31": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_98": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_32": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_99": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_176": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_100": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_138": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_43": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_42": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_139": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_62": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_177": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_140": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_178": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_32": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_101": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_58": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_102": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_85": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_103": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_104": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_23": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_120": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_10": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_27": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_10": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_121": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_33": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_122": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_59": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_123": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_124": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_125": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_86": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_87": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_88": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_13": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_10": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_16": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_141": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_10": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_179": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_180": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_21": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_142": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_63": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_181": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_182": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_22": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_143": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_64": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_183": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_10": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_126": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_127": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_60": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_22": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_89": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_90": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_44": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_43": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_144": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_65": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_184": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_185": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_34": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_186": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_91": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_105": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_145": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_45": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_44": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_146": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_66": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_187": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_147": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_188": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_33": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_106": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_35": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_107": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_189": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_108": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_148": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_46": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_45": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_149": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_67": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_190": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_150": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_191": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_34": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_109": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_61": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_110": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_92": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_111": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_112": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_24": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_128": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_11": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_28": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_11": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_129": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_36": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_130": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_62": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_131": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_132": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_133": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_93": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_94": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_95": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_14": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_11": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_17": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_151": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_11": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_192": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_193": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_23": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_152": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_68": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_194": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_195": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_24": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_153": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_69": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_196": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_11": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_134": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_135": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_63": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_23": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_96": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_97": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_47": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_46": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_154": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_70": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_197": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_198": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_37": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_199": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_98": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_113": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_155": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_48": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_47": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_156": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_71": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_200": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_157": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_201": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_35": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_114": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_38": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_115": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_202": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_116": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_158": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_49": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_48": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_159": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_72": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_203": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_160": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_204": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_36": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_117": { + "shape": [ + 1, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "view_99": { + "shape": [ + 1, + 16, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "slice_136": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_137": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_24": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "_unsafe_view_9": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_100": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_50": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_49": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "add_161": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "rsqrt_73": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "mul_205": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_162": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_206": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "type_as_37": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_64": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_118": { + "shape": [ + 1, + 16, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_101": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_51": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_50": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "add_163": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "rsqrt_74": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "mul_207": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_164": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_208": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_38": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_65": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_119": { + "shape": [ + 1, + 16, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_102": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_66": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "unsqueeze_29": { + "shape": [ + 1, + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "unsqueeze_30": { + "shape": [ + 1, + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_138": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_139": { + "shape": [ + 1, + 8, + 16, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_140": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_141": { + "shape": [ + 1, + 2, + 16, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "mul_209": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_142": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_143": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "neg_18": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "cat_25": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_210": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_165": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_211": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_144": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_145": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "neg_19": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_26": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_212": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_166": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "cat_27": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "cat_28": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cache_update_6": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_7": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_31": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_8": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_25": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_10": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_32": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_9": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_26": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_11": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_67": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_7": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "mul_213": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "add_167": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "softmax_3": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "matmul_8": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_68": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_27": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_103": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_15": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_214": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_120": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_168": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_52": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_51": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_169": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_75": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_215": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_170": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_216": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_39": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_121": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_39": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_122": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_217": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_123": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_171": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_53": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_52": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_172": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_76": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_218": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_173": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_219": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_40": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_124": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_69": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_125": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_104": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_126": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_127": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_29": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_146": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_12": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_33": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_12": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_147": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_40": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_148": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_70": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_149": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_150": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_151": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_105": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_106": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_107": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_16": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_12": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_20": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_174": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_12": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_220": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_221": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_25": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_175": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_77": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_222": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_223": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_26": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_176": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_78": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_224": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_12": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_152": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_153": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_71": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_28": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_108": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_109": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_54": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_53": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_177": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_79": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_225": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_226": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_41": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_227": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_110": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_128": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_178": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_55": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_54": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_179": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_80": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_228": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_180": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_229": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_41": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_129": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_42": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_130": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_230": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_131": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_181": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_56": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_55": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_182": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_81": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_231": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_183": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_232": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_42": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_132": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_72": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_133": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_111": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_134": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_135": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_30": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_154": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_13": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_34": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_13": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_155": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_43": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_156": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_73": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_157": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_158": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_159": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_112": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_113": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_114": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_17": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_13": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_21": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_184": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_13": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_233": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_234": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_27": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_185": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_82": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_235": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_236": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_28": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_186": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_83": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_237": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_13": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_160": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_161": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_74": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_29": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_115": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_116": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_57": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_56": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_187": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_84": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_238": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_239": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_44": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_240": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_117": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_136": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_188": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_58": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_57": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_189": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_85": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_241": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_190": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_242": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_43": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_137": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_45": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_138": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_243": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_139": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_191": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_59": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_58": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_192": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_86": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_244": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_193": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_245": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_44": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_140": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_75": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_141": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_118": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_142": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_143": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_31": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_162": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_14": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_35": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_14": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_163": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_46": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_164": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_76": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_165": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_166": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_167": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_119": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_120": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_121": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_18": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_14": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_22": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_194": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_14": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_246": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_247": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_29": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_195": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_87": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_248": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_249": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_30": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_196": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_88": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_250": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_14": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_168": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_169": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_77": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_30": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_122": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_123": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_60": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_59": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_197": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_89": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_251": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_252": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_47": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_253": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_124": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_144": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_198": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_61": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_60": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_199": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_90": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_254": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_200": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_255": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_45": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_145": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_48": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_146": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_256": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_147": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_201": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_62": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_61": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_202": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_91": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_257": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_203": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_258": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_46": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_148": { + "shape": [ + 1, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "view_125": { + "shape": [ + 1, + 16, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "slice_170": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_171": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_31": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "_unsafe_view_12": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_126": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_63": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_62": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "add_204": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "rsqrt_92": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "mul_259": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_205": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_260": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "type_as_47": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_78": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_149": { + "shape": [ + 1, + 16, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_127": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_64": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_63": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "add_206": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "rsqrt_93": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "mul_261": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_207": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_262": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_48": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_79": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_150": { + "shape": [ + 1, + 16, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_128": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_80": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "unsqueeze_36": { + "shape": [ + 1, + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "unsqueeze_37": { + "shape": [ + 1, + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_172": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_173": { + "shape": [ + 1, + 8, + 16, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_174": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_175": { + "shape": [ + 1, + 2, + 16, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "mul_263": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_176": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_177": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "neg_23": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "cat_32": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_264": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_208": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_265": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_178": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_179": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "neg_24": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_33": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_266": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_209": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "cat_34": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "cat_35": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cache_update_8": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_9": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_38": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_10": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_32": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_13": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_39": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_11": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_33": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_14": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_81": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_9": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "mul_267": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "add_210": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "softmax_4": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "matmul_10": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_82": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_34": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_129": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_19": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_268": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_151": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_211": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_65": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_64": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_212": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_94": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_269": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_213": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_270": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_49": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_152": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_49": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_153": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_271": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_154": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_214": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_66": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_65": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_215": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_95": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_272": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_216": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_273": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_50": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_155": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_83": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_156": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_130": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_157": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_158": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_36": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_180": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_15": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_40": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_15": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_181": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_50": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_182": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_84": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_183": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_184": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_185": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_131": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_132": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_133": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_20": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_15": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_25": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_217": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_15": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_274": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_275": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_31": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_218": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_96": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_276": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_277": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_32": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_219": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_97": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_278": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_15": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_186": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_187": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_85": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_35": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_134": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_135": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_67": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_66": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_220": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_98": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_279": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_280": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_51": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_281": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_136": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_159": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_221": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_68": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_67": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_222": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_99": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_282": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_223": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_283": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_51": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_160": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_52": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_161": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_284": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_162": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_224": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_69": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_68": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_225": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_100": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_285": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_226": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_286": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_52": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_163": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_86": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_164": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_137": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_165": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_166": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_37": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_188": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_16": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_41": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_16": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_189": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_53": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_190": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_87": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_191": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_192": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_193": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_138": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_139": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_140": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_21": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_16": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_26": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_227": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_16": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_287": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_288": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_33": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_228": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_101": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_289": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_290": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_34": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_229": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_102": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_291": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_16": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_194": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_195": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_88": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_36": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_141": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_142": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_70": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_69": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_230": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_103": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_292": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_293": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_54": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_294": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_143": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_167": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_231": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_71": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_70": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_232": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_104": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_295": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_233": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_296": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_53": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_168": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_55": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_169": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_297": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_170": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_234": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_72": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_71": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_235": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_105": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_298": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_236": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_299": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_54": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_171": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_89": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "linear_172": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_144": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_173": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "linear_174": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_38": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_196": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_17": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_42": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_17": { + "shape": [ + 1, + 6144, + 23 + ], + "dtype": "float32", + "role": "value", + "byteLength": 565248 + }, + "slice_197": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "silu_56": { + "shape": [ + 1, + 6144, + 20 + ], + "dtype": "float32", + "role": "value", + "byteLength": 491520 + }, + "slice_198": { + "shape": [ + 1, + 6144, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "transpose_90": { + "shape": [ + 1, + 16, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_199": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_200": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_201": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_145": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_146": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_147": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_22": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "exp_17": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_27": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_237": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "softplus_17": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_300": { + "shape": [ + 1, + 16, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_301": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_35": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_238": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_106": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_302": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_303": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sum_36": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_239": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_107": { + "shape": [ + 1, + 16, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_304": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "normalized_gated_delta_rule_17": { + "shape": [ + 1, + 16, + 144, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1179648 + }, + "slice_202": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_203": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_91": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_37": { + "shape": [ + 1, + 16, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_148": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_149": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_73": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_72": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "add_240": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "rsqrt_108": { + "shape": [ + 256, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_305": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_306": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "silu_57": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_307": { + "shape": [ + 256, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_150": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_175": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_241": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_74": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_73": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_242": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_109": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_308": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_243": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_309": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_55": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_176": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_58": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_177": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_310": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_178": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_244": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_75": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_74": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_245": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_110": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_311": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_246": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_312": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_56": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_179": { + "shape": [ + 1, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "view_151": { + "shape": [ + 1, + 16, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "slice_204": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_205": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_38": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "_unsafe_view_15": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_152": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_76": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_75": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "add_247": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "rsqrt_111": { + "shape": [ + 1, + 16, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "mul_313": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_248": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_314": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "type_as_57": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_92": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_180": { + "shape": [ + 1, + 16, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_153": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_77": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_76": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "add_249": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "rsqrt_112": { + "shape": [ + 1, + 16, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "mul_315": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_250": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_316": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_58": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_93": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_181": { + "shape": [ + 1, + 16, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_154": { + "shape": [ + 1, + 16, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_94": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "unsqueeze_43": { + "shape": [ + 1, + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "unsqueeze_44": { + "shape": [ + 1, + 1, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_206": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_207": { + "shape": [ + 1, + 8, + 16, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_208": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_209": { + "shape": [ + 1, + 2, + 16, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "mul_317": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_210": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_211": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "neg_28": { + "shape": [ + 1, + 8, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "cat_39": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_318": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_251": { + "shape": [ + 1, + 8, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_319": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_212": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_213": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "neg_29": { + "shape": [ + 1, + 2, + 16, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_40": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_320": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_252": { + "shape": [ + 1, + 2, + 16, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "cat_41": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "cat_42": { + "shape": [ + 1, + 2, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cache_update_10": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_11": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_45": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_12": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_39": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_16": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_46": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_13": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_40": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_17": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_95": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_11": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "mul_321": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "add_253": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "softmax_5": { + "shape": [ + 1, + 8, + 16, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "matmul_12": { + "shape": [ + 1, + 8, + 16, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_96": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "clone_41": { + "shape": [ + 1, + 16, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_155": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "sigmoid_23": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_322": { + "shape": [ + 1, + 16, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_182": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_254": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_78": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_77": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_255": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_113": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_323": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_256": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_324": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_59": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_183": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "silu_59": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_184": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "mul_325": { + "shape": [ + 1, + 16, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 229376 + }, + "linear_185": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_257": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "pow_79": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "mean_78": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_258": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "rsqrt_114": { + "shape": [ + 1, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "mul_326": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "add_259": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_327": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "type_as_60": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "alias": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_186": { + "shape": [ + 1, + 16, + 248320 + ], + "dtype": "float32", + "role": "value", + "byteLength": 15892480 + }, + "p_model_model_language_model_layers_0_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 3632512, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 4091264, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 4550016, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 5008768, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 5467520, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 5926272, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 6385024, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 6843776, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 7302528, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 7761280, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 8220032, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 8678784, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 9137536, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 9596288, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 10055040, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 10513792, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 10972544, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 11431296, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 11890048, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 12348800, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 12807552, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 13266304, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 13725056, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 14183808, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 14642560, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 15101312, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 15560064, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 16018816, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 16477568, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 16936320, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 17395072, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 17853824, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 18312576, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 18771328, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 19230080, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 19688832, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 20147584, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 20606336, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 21065088, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 21523840, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 21982592, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 22441344, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 22900096, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 23358848, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 23817600, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 24276352, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 24735104, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 25193856, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 25652608, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 26111360, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 26570112, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 27028864, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 27487616, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 27946368, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 28405120, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 28863872, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 29322624, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 29781376, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 30240128, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 30698880, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 31157632, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 31616384, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 32075136, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 32533888, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 32992640, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 33451392, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 33910144, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 34368896, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 34827648, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 35286400, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 35745152, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 36203904, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 36662656, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 36664704, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 36666752, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37453184, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37715328, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37977472, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37979520, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37981568, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 38768000, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 39030144, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 39292288, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 39357824, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 39619968, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 40144256, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 40209792, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 40211840, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 40213888, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41000320, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41262464, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41524608, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41526656, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41528704, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42315136, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42577280, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42839424, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42841472, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42843520, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 43629952, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 43892096, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 44154240, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 44219776, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 44481920, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45006208, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45071744, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45073792, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45075840, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45862272, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 46124416, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 46386560, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 46388608, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 46390656, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47177088, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47439232, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47701376, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47703424, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47705472, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 48491904, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 48754048, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49016192, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49081728, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49343872, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49868160, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49933696, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49935744, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49937792, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 50724224, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 50986368, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 51248512, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 51250560, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 51252608, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52039040, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52301184, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52563328, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52565376, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52567424, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53353856, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53616000, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53878144, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53880192, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53882240, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 54668672, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 54930816, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 55192960, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 55258496, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 55520640, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56044928, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56110464, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56112512, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56114560, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56900992, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 57163136, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 57425280, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 57490816, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 57752960, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 58277248, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 58342784, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 58344832, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 58346880, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59133312, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59395456, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59657600, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59659648, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59661696, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60448128, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60710272, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60972416, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60974464, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60976512, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 61762944, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 62025088, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 62287232, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 62352768, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 62614912, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63139200, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63204736, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63206784, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63208832, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63995264, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 64257408, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 64519552, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 64521600, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 64523648, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 65310080, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 65572224, + "byteLength": 262144 + } + ] + }, + "embedding__part_0": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "embedding__part_1": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "embedding__part_2": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "embedding__part_3": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "embedding__sum_0": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "embedding__sum_1": { + "shape": [ + 1, + 16, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "linear_186__part_0": { + "shape": [ + 1, + 16, + 131072 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "linear_186__part_1": { + "shape": [ + 1, + 16, + 117248 + ], + "dtype": "float32", + "role": "value", + "byteLength": 7503872 + }, + "linear_186__part_2": { + "shape": [ + 1, + 16, + 65536 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4194304 + }, + "linear_186__part_3": { + "shape": [ + 1, + 16, + 51712 + ], + "dtype": "float32", + "role": "value", + "byteLength": 3309568 + }, + "p_model_model_language_model_embed_tokens_weight__int8_scales": { + "shape": [ + 248320, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 31784960, + "parts": [ + { + "path": "weights/int8-422da3921400-scales-000.bin", + "offset": 0, + "byteLength": 31784960 + } + ] + }, + "p_model_model_language_model_embed_tokens_weight__rows_0_131072": { + "shape": [ + 131072, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 134217728, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 0 + }, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/int8-422da3921400-000.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-001.bin", + "offset": 0, + "byteLength": 67108864 + } + ] + }, + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072": { + "shape": [ + 131072, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16777216, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight__int8_scales", + "offset": 0 + }, + "parts": [ + { + "path": "weights/int8-422da3921400-scales-000.bin", + "offset": 0, + "byteLength": 16777216 + } + ] + }, + "p_model_model_language_model_embed_tokens_weight__rows_131072_117248": { + "shape": [ + 117248, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 120061952, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 134217728 + }, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/int8-422da3921400-002.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-003.bin", + "offset": 0, + "byteLength": 52953088 + } + ] + }, + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248": { + "shape": [ + 117248, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 15007744, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight__int8_scales", + "offset": 16777216 + }, + "parts": [ + { + "path": "weights/int8-422da3921400-scales-000.bin", + "offset": 16777216, + "byteLength": 15007744 + } + ] + } + }, + "nodes": [ + { + "id": "embedding", + "op": "aten.embedding.default", + "args": [ + { + "tensor": "p_model_model_language_model_embed_tokens_weight" + }, + { + "tensor": "input_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1527, in forward\n inputs_embeds = self.get_input_embeddings()(input_ids)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "position_ids" + }, + 0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1179, in forward\n position_ids = position_ids[None, ...].expand(4, position_ids.shape[0], -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze" + }, + [ + 4, + 1, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1179, in forward\n position_ids = position_ids[None, ...].expand(4, position_ids.shape[0], -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_1", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "expand" + }, + 0, + 1, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1183, in forward\n position_ids = position_ids[1:]\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_1", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "b_model_model_language_model_rotary_emb_inv_freq" + }, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 5, in forward\n unsqueeze_1 = torch.ops.aten.unsqueeze.default(b_model_model_language_model_rotary_emb_inv_freq, 0); b_model_model_language_model_rotary_emb_inv_freq = None" + }, + { + "id": "unsqueeze_2", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "unsqueeze_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \".49\", line 6, in forward\n unsqueeze_2 = torch.ops.aten.unsqueeze.default(unsqueeze_1, 1); unsqueeze_1 = None" + }, + { + "id": "unsqueeze_3", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "unsqueeze_2" + }, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 7, in forward\n unsqueeze_3 = torch.ops.aten.unsqueeze.default(unsqueeze_2, 3); unsqueeze_2 = None" + }, + { + "id": "expand_1", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_3" + }, + [ + 3, + 1, + -1, + 1 + ] + ], + "kwargs": {}, + "source": "File \".49\", line 10, in forward\n expand_1 = torch.ops.aten.expand.default(to, [3, 1, -1, 1]); to = None" + }, + { + "id": "unsqueeze_4", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "slice_1" + }, + 2 + ], + "kwargs": {}, + "source": "File \".49\", line 13, in forward\n unsqueeze_4 = torch.ops.aten.unsqueeze.default(slice_1, 2); slice_1 = None" + }, + { + "id": "_to_copy", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "unsqueeze_4" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \".49\", line 15, in forward\n to_2 = torch.ops.aten.to.dtype(unsqueeze_4, torch.float32); unsqueeze_4 = None" + }, + { + "id": "matmul", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "expand_1" + }, + { + "tensor": "_to_copy" + } + ], + "kwargs": {}, + "source": "File \".49\", line 20, in forward\n matmul = torch.ops.aten.matmul.default(to_3, to_4); to_3 = to_4 = None" + }, + { + "id": "transpose", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 21, in forward\n transpose = torch.ops.aten.transpose.int(matmul, 2, 3); matmul = None" + }, + { + "id": "select", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 22, in forward\n select = torch.ops.aten.select.int(transpose, 0, 0)" + }, + { + "id": "select_1", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose" + }, + 0, + 1 + ], + "kwargs": {}, + "source": "File \".49\", line 23, in forward\n select_1 = torch.ops.aten.select.int(transpose, 0, 1)" + }, + { + "id": "slice_2", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "select_1" + }, + 2, + 1, + 33, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 24, in forward\n slice_2 = torch.ops.aten.slice.Tensor(select_1, 2, 1, 33, 3); select_1 = None" + }, + { + "id": "slice_3", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "select" + }, + 2, + 1, + 33, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 25, in forward\n slice_3 = torch.ops.aten.slice.Tensor(select, 2, 1, 33, 3)" + }, + { + "id": "copy", + "op": "aten.copy.default", + "args": [ + { + "tensor": "slice_3" + }, + { + "tensor": "slice_2" + } + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "transpose_1", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "select_2", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_1" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "slice_scatter", + "op": "aten.slice_scatter.default", + "args": [ + { + "tensor": "select_2" + }, + { + "tensor": "copy" + }, + 2, + 1, + 33, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "select_scatter", + "op": "aten.select_scatter.default", + "args": [ + { + "tensor": "transpose_1" + }, + { + "tensor": "slice_scatter" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "transpose_2", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "select_scatter" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "transpose_7", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "transpose_2" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 29, in forward\n slice_5 = torch.ops.aten.slice.Tensor(select, 2, 2, 30, 3)" + }, + { + "id": "select_7", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_7" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 29, in forward\n slice_5 = torch.ops.aten.slice.Tensor(select, 2, 2, 30, 3)" + }, + { + "id": "slice_7", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "select_7" + }, + 2, + 2, + 30, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 29, in forward\n slice_5 = torch.ops.aten.slice.Tensor(select, 2, 2, 30, 3)" + }, + { + "id": "transpose_8", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "transpose_2" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 28, in forward\n slice_4 = torch.ops.aten.slice.Tensor(select_2, 2, 2, 30, 3); select_2 = None" + }, + { + "id": "select_8", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_8" + }, + 0, + 2 + ], + "kwargs": {}, + "source": "File \".49\", line 28, in forward\n slice_4 = torch.ops.aten.slice.Tensor(select_2, 2, 2, 30, 3); select_2 = None" + }, + { + "id": "slice_8", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "select_8" + }, + 2, + 2, + 30, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 28, in forward\n slice_4 = torch.ops.aten.slice.Tensor(select_2, 2, 2, 30, 3); select_2 = None" + }, + { + "id": "copy_1", + "op": "aten.copy.default", + "args": [ + { + "tensor": "slice_7" + }, + { + "tensor": "slice_8" + } + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "transpose_9", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "transpose_2" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "select_9", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_9" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "slice_scatter_1", + "op": "aten.slice_scatter.default", + "args": [ + { + "tensor": "select_9" + }, + { + "tensor": "copy_1" + }, + 2, + 2, + 30, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "select_scatter_1", + "op": "aten.select_scatter.default", + "args": [ + { + "tensor": "transpose_9" + }, + { + "tensor": "slice_scatter_1" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "transpose_10", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "select_scatter_1" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "transpose_12", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "transpose_10" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 22, in forward\n select = torch.ops.aten.select.int(transpose, 0, 0)" + }, + { + "id": "select_11", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_12" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 22, in forward\n select = torch.ops.aten.select.int(transpose, 0, 0)" + }, + { + "id": "cat", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "select_11" + }, + { + "tensor": "select_11" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \".49\", line 31, in forward\n cat = torch.ops.aten.cat.default([select, select], -1); select = None" + }, + { + "id": "cos", + "op": "aten.cos.default", + "args": [ + { + "tensor": "cat" + } + ], + "kwargs": {}, + "source": "File \".49\", line 32, in forward\n cos = torch.ops.aten.cos.default(cat)" + }, + { + "id": "mul", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cos" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \".49\", line 33, in forward\n mul = torch.ops.aten.mul.Tensor(cos, 1.0); cos = None" + }, + { + "id": "sin", + "op": "aten.sin.default", + "args": [ + { + "tensor": "cat" + } + ], + "kwargs": {}, + "source": "File \".49\", line 34, in forward\n sin = torch.ops.aten.sin.default(cat); cat = None" + }, + { + "id": "mul_1", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "sin" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \".49\", line 35, in forward\n mul_1 = torch.ops.aten.mul.Tensor(sin, 1.0); sin = None" + }, + { + "id": "pow_1", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "embedding" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_1" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_2", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "embedding" + }, + { + "tensor": "rsqrt" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_1", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_3", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_2" + }, + { + "tensor": "add_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_3" + }, + { + "tensor": "embedding" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_13", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_1", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_1" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_2", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_3", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_1", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_0" + }, + { + "tensor": "transpose_13" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_10", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_1" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_5", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_1" + }, + { + "tensor": "unsqueeze_5" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_11", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_12", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_14", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_12" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_13", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_14" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_14", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_14" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_15", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_14" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_1", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_13" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_2", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_14" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_3", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_15" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_2", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_3" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_4", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg" + }, + { + "tensor": "softplus" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_5", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_1" + }, + { + "tensor": "view_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_1", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_5" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_3", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_1" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_1", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_6", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_1" + }, + { + "tensor": "rsqrt_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_7", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_2" + }, + { + "tensor": "view_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_2", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_7" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_4", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_2" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_2", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_8", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_2" + }, + { + "tensor": "rsqrt_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_6" + }, + { + "tensor": "mul_8" + }, + { + "tensor": "view_3" + }, + { + "tensor": "mul_4" + }, + { + "tensor": "sigmoid" + }, + { + "tensor": "states_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_16", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_17", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_15", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_17" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_15" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_4", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_5", + "op": "aten.view.default", + "args": [ + { + "tensor": "view" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_2", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_4" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_1", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_2" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_5", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_1" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_3", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_9", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_4" + }, + { + "tensor": "rsqrt_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_10", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_norm_weight" + }, + { + "tensor": "mul_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_1", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_11", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_10" + }, + { + "tensor": "silu_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_6", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_11" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_4", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_6" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_6", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "embedding" + }, + { + "tensor": "linear_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_3", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_6" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_2", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_3" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_7", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_2" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_4", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_12", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_6" + }, + { + "tensor": "rsqrt_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_8", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_13", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_12" + }, + { + "tensor": "add_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_1", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_13" + }, + { + "tensor": "add_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_5", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_1" + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_2", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_6", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_1" + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_14", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_2" + }, + { + "tensor": "linear_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_7", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_14" + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_9", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_6" + }, + { + "tensor": "linear_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_4", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_9" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_3", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_4" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_10", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_3" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_5", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_15", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_9" + }, + { + "tensor": "rsqrt_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_11", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_16", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_15" + }, + { + "tensor": "add_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_2", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_16" + }, + { + "tensor": "add_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_8", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_2" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_16", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_8" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_9", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_2" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_7", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_9" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_10", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_2" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_11", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_2" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_2", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_2" + }, + { + "tensor": "transpose_16" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_18", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_2" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_1", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_6", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_1", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_2" + }, + { + "tensor": "unsqueeze_6" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_19", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_1" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_3", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_20", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_3" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_17", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_20" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_21", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_17" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_22", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_17" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_23", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_17" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_8", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_21" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_9", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_22" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_10", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_23" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_1", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_1", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_1", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_12", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_11" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_1", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_17", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_1" + }, + { + "tensor": "softplus_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_18", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_8" + }, + { + "tensor": "view_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_3", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_18" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_13", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_3" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_6", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_19", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_8" + }, + { + "tensor": "rsqrt_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_20", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_9" + }, + { + "tensor": "view_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_4", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_20" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_14", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_4" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_7", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_21", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_9" + }, + { + "tensor": "rsqrt_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_1", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_19" + }, + { + "tensor": "mul_21" + }, + { + "tensor": "view_10" + }, + { + "tensor": "mul_17" + }, + { + "tensor": "sigmoid_1" + }, + { + "tensor": "states_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_24", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_1" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_25", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_1" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_18", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_25" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_1", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_18" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_11", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_1" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_12", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_7" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_5", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_11" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_4", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_5" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_15", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_4" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_8", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_22", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_11" + }, + { + "tensor": "rsqrt_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_23", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_norm_weight" + }, + { + "tensor": "mul_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_4", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_24", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_23" + }, + { + "tensor": "silu_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_13", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_24" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_12", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_13" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_16", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_9" + }, + { + "tensor": "linear_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_6", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_16" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_5", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_6" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_17", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_5" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_9", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_25", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_16" + }, + { + "tensor": "rsqrt_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_18", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_26", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_25" + }, + { + "tensor": "add_18" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_3", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_26" + }, + { + "tensor": "add_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_13", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_3" + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_5", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_14", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_3" + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_27", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_5" + }, + { + "tensor": "linear_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_15", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_27" + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_19", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_16" + }, + { + "tensor": "linear_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_7", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_19" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_6", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_7" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_20", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_6" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_10", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_20" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_28", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_19" + }, + { + "tensor": "rsqrt_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_21", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_29", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_28" + }, + { + "tensor": "add_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_4", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_29" + }, + { + "tensor": "add_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_16", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_4" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_19", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_16" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_17", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_4" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_14", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_17" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_18", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_4" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_19", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_4" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_3", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_4" + }, + { + "tensor": "transpose_19" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_26", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_3" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_2", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_7", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_2" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_2", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_3" + }, + { + "tensor": "unsqueeze_7" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_27", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_2" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_6", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_28", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_6" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_20", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_28" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_29", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_20" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_30", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_20" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_31", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_20" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_15", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_29" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_16", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_30" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_17", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_31" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_2", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_18" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_2", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_2", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_22", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_19" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_2", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_30", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_2" + }, + { + "tensor": "softplus_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_31", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_15" + }, + { + "tensor": "view_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_5", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_31" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_23", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_5" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_11", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_32", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_15" + }, + { + "tensor": "rsqrt_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_33", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_16" + }, + { + "tensor": "view_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_6", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_33" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_24", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_6" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_12", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_24" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_34", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_16" + }, + { + "tensor": "rsqrt_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_2", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_32" + }, + { + "tensor": "mul_34" + }, + { + "tensor": "view_17" + }, + { + "tensor": "mul_30" + }, + { + "tensor": "sigmoid_2" + }, + { + "tensor": "states_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_32", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_2" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_33", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_2" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_21", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_33" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_2", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_21" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_18", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_2" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_19", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_14" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_8", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_18" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_7", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_8" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_25", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_7" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_13", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_25" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_35", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_18" + }, + { + "tensor": "rsqrt_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_36", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_norm_weight" + }, + { + "tensor": "mul_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_7", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_37", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_36" + }, + { + "tensor": "silu_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_20", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_37" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_20", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_20" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_26", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_19" + }, + { + "tensor": "linear_20" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_9", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_26" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_8", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_9" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_27", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_8" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_14", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_38", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_26" + }, + { + "tensor": "rsqrt_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_28", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_39", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_38" + }, + { + "tensor": "add_28" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_5", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_39" + }, + { + "tensor": "add_26" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_21", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_5" + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_8", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_22", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_5" + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_40", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_8" + }, + { + "tensor": "linear_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_23", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_40" + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_29", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_26" + }, + { + "tensor": "linear_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_10", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_29" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_9", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_10" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_30", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_9" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_15", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_41", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_29" + }, + { + "tensor": "rsqrt_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_31", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_42", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_41" + }, + { + "tensor": "add_31" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_6", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_42" + }, + { + "tensor": "add_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_24", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_6" + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_21", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_24" + }, + [ + 1, + 16, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_34", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_21" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_35", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_21" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_3", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_35" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_3" + }, + [ + 1, + 16, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_22", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_34" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_11", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_22" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_10", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_11" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_32", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_10" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_16", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_43", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_22" + }, + { + "tensor": "rsqrt_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_33", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_44", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_43" + }, + { + "tensor": "add_33" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_7", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_44" + }, + { + "tensor": "view_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_22", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_7" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_25", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_6" + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_23", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_25" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_12", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_23" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_11", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_12" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_34", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_11" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_17", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_34" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_45", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_23" + }, + { + "tensor": "rsqrt_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_35", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_46", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_45" + }, + { + "tensor": "add_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_8", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_46" + }, + { + "tensor": "view_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_23", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_8" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_26", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_6" + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_24", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_26" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_24", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_24" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_8", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_9", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_36", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_22" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_37", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_22" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_38", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_23" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_39", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_23" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_47", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_36" + }, + { + "tensor": "unsqueeze_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_40", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_36" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_41", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_36" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_3", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_4", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_3" + }, + { + "tensor": "slice_40" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_48", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_4" + }, + { + "tensor": "unsqueeze_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_36", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_47" + }, + { + "tensor": "mul_48" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_49", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_38" + }, + { + "tensor": "unsqueeze_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_42", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_38" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_43", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_38" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_4", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_5", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_4" + }, + { + "tensor": "slice_42" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_50", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_5" + }, + { + "tensor": "unsqueeze_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_37", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_49" + }, + { + "tensor": "mul_50" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_6", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_36" + }, + { + "tensor": "slice_37" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_7", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_37" + }, + { + "tensor": "slice_39" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_6" + }, + { + "tensor": "cat_7" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_1", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_7" + }, + { + "tensor": "transpose_24" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_10", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_2", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_10" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_4", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_2" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_1", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_4" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_11", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_1" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_3", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_11" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_5", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_3" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_2", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_5" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_25", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_1" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_1", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_6" + }, + { + "tensor": "transpose_25" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_51", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_1" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_38", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_51" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_38" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_2", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax" + }, + { + "tensor": "_unsafe_view_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_26", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_2" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_6", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_26" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_25", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_6" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_3", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_52", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_25" + }, + { + "tensor": "sigmoid_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_27", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_52" + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_39", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_29" + }, + { + "tensor": "linear_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_13", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_39" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_12", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_13" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_40", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_12" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_18", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_40" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_53", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_39" + }, + { + "tensor": "rsqrt_18" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_41", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_54", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_53" + }, + { + "tensor": "add_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_9", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_54" + }, + { + "tensor": "add_39" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_28", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_9" + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_9", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_28" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_29", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_9" + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_55", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_9" + }, + { + "tensor": "linear_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_30", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_55" + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_42", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_39" + }, + { + "tensor": "linear_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_14", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_42" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_13", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_14" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_43", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_13" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_19", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_56", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_42" + }, + { + "tensor": "rsqrt_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_44", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_57", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_56" + }, + { + "tensor": "add_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_10", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_57" + }, + { + "tensor": "add_42" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_31", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_10" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_27", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_31" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_32", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_10" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_26", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_32" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_33", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_10" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_34", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_10" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_8", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_8" + }, + { + "tensor": "transpose_27" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_44", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_8" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_3", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_12", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_3" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_3", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_8" + }, + { + "tensor": "unsqueeze_12" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_45", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_3" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_10", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_46", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_10" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_28", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_46" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_47", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_28" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_48", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_28" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_49", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_28" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_27", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_47" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_28", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_48" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_29", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_49" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_4", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_33" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_3", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_5", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_45", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_34" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_3", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_58", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_5" + }, + { + "tensor": "softplus_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_59", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_27" + }, + { + "tensor": "view_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_7", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_59" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_46", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_7" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_20", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_46" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_60", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_27" + }, + { + "tensor": "rsqrt_20" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_61", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_28" + }, + { + "tensor": "view_28" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_8", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_61" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_47", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_8" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_21", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_47" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_62", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_28" + }, + { + "tensor": "rsqrt_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_3", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_60" + }, + { + "tensor": "mul_62" + }, + { + "tensor": "view_29" + }, + { + "tensor": "mul_58" + }, + { + "tensor": "sigmoid_4" + }, + { + "tensor": "states_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_50", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_3" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_51", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_3" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_29", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_51" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_7", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_29" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_30", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_7" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_31", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_26" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_15", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_30" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_14", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_15" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_48", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_14" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_22", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_48" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_63", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_30" + }, + { + "tensor": "rsqrt_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_64", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_norm_weight" + }, + { + "tensor": "mul_63" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_11", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_31" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_65", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_64" + }, + { + "tensor": "silu_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_32", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_65" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_35", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_32" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_49", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_42" + }, + { + "tensor": "linear_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_16", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_49" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_15", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_16" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_50", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_15" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_23", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_50" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_66", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_49" + }, + { + "tensor": "rsqrt_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_51", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_67", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_66" + }, + { + "tensor": "add_51" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_11", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_67" + }, + { + "tensor": "add_49" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_36", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_11" + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_12", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_36" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_37", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_11" + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_68", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_12" + }, + { + "tensor": "linear_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_38", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_68" + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_52", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_49" + }, + { + "tensor": "linear_38" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_17", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_52" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_16", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_17" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_53", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_16" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_24", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_69", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_52" + }, + { + "tensor": "rsqrt_24" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_54", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_70", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_69" + }, + { + "tensor": "add_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_12", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_70" + }, + { + "tensor": "add_52" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_39", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_12" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_30", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_39" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_40", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_12" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_33", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_40" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_41", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_12" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_42", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_12" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_9", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_10" + }, + { + "tensor": "transpose_30" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_52", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_9" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_4", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_13", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_4" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_4", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_9" + }, + { + "tensor": "unsqueeze_13" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_53", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_4" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_13", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_54", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_13" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_31", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_54" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_55", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_31" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_56", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_31" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_57", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_31" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_34", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_55" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_35", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_56" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_36", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_57" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_5", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_4", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_6", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_55", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_42" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_4", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_55" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_71", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_6" + }, + { + "tensor": "softplus_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_72", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_34" + }, + { + "tensor": "view_34" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_9", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_72" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_56", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_9" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_25", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_56" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_73", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_34" + }, + { + "tensor": "rsqrt_25" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_74", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_35" + }, + { + "tensor": "view_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_10", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_74" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_57", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_10" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_26", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_57" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_75", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_35" + }, + { + "tensor": "rsqrt_26" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_4", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_73" + }, + { + "tensor": "mul_75" + }, + { + "tensor": "view_36" + }, + { + "tensor": "mul_71" + }, + { + "tensor": "sigmoid_5" + }, + { + "tensor": "states_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_58", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_4" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_59", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_4" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_32", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_59" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_8", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_32" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_37", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_8" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_38", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_33" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_18", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_37" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_17", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_18" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_58", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_17" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_27", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_58" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_76", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_37" + }, + { + "tensor": "rsqrt_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_77", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_norm_weight" + }, + { + "tensor": "mul_76" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_14", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_38" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_78", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_77" + }, + { + "tensor": "silu_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_39", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_78" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_43", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_39" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_59", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_52" + }, + { + "tensor": "linear_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_19", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_59" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_18", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_19" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_60", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_18" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_28", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_79", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_59" + }, + { + "tensor": "rsqrt_28" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_61", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_80", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_79" + }, + { + "tensor": "add_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_13", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_80" + }, + { + "tensor": "add_59" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_44", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_13" + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_15", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_45", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_13" + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_81", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_15" + }, + { + "tensor": "linear_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_46", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_81" + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_62", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_59" + }, + { + "tensor": "linear_46" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_20", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_62" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_19", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_20" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_63", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_19" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_29", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_63" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_82", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_62" + }, + { + "tensor": "rsqrt_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_64", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_83", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_82" + }, + { + "tensor": "add_64" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_14", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_83" + }, + { + "tensor": "add_62" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_47", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_14" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_33", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_47" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_48", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_14" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_40", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_48" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_49", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_14" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_50", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_14" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_10", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_12" + }, + { + "tensor": "transpose_33" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_60", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_10" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_5", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_14", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_5" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_5", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_10" + }, + { + "tensor": "unsqueeze_14" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_61", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_5" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_16", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_62", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_16" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_34", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_62" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_63", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_34" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_64", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_34" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_65", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_34" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_41", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_63" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_42", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_64" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_43", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_65" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_6", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_49" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_5", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_7", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_65", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_50" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_5", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_65" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_84", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_7" + }, + { + "tensor": "softplus_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_85", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_41" + }, + { + "tensor": "view_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_11", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_85" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_66", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_11" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_30", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_66" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_86", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_41" + }, + { + "tensor": "rsqrt_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_87", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_42" + }, + { + "tensor": "view_42" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_12", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_87" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_67", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_12" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_31", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_88", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_42" + }, + { + "tensor": "rsqrt_31" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_5", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_86" + }, + { + "tensor": "mul_88" + }, + { + "tensor": "view_43" + }, + { + "tensor": "mul_84" + }, + { + "tensor": "sigmoid_6" + }, + { + "tensor": "states_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_66", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_5" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_67", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_5" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_35", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_67" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_9", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_35" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_44", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_9" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_45", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_40" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_21", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_44" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_20", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_21" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_68", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_20" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_32", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_68" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_89", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_44" + }, + { + "tensor": "rsqrt_32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_90", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_norm_weight" + }, + { + "tensor": "mul_89" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_17", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_91", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_90" + }, + { + "tensor": "silu_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_46", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_91" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_51", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_46" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_69", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_62" + }, + { + "tensor": "linear_51" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_22", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_69" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_21", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_22" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_70", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_21" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_33", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_70" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_92", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_69" + }, + { + "tensor": "rsqrt_33" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_71", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_93", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_92" + }, + { + "tensor": "add_71" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_15", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_93" + }, + { + "tensor": "add_69" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_52", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_15" + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_18", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_52" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_53", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_15" + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_94", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_18" + }, + { + "tensor": "linear_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_54", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_94" + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_72", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_69" + }, + { + "tensor": "linear_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_23", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_72" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_22", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_23" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_73", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_22" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_34", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_73" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_95", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_72" + }, + { + "tensor": "rsqrt_34" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_74", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_96", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_95" + }, + { + "tensor": "add_74" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_16", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_96" + }, + { + "tensor": "add_72" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_55", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_16" + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_47", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_55" + }, + [ + 1, + 16, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_68", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_47" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_69", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_47" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_10", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_69" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_3", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_10" + }, + [ + 1, + 16, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_48", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_68" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_24", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_48" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_23", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_24" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_75", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_23" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_35", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_97", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_48" + }, + { + "tensor": "rsqrt_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_76", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_98", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_97" + }, + { + "tensor": "add_76" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_17", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_98" + }, + { + "tensor": "view_48" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_36", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_17" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_56", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_16" + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_49", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_56" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_25", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_49" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_24", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_25" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_77", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_24" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_36", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_77" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_99", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_49" + }, + { + "tensor": "rsqrt_36" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_78", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_100", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_99" + }, + { + "tensor": "add_78" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_18", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_100" + }, + { + "tensor": "view_49" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_37", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_18" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_57", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_16" + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_50", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_57" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_38", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_50" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_15", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_16", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_70", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_36" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_71", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_36" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_72", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_37" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_73", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_37" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_101", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_70" + }, + { + "tensor": "unsqueeze_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_74", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_70" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_75", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_70" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_8", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_11", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_8" + }, + { + "tensor": "slice_74" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_102", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_11" + }, + { + "tensor": "unsqueeze_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_79", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_101" + }, + { + "tensor": "mul_102" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_103", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_72" + }, + { + "tensor": "unsqueeze_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_76", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_72" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_77", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_72" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_9", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_77" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_12", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_9" + }, + { + "tensor": "slice_76" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_104", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_12" + }, + { + "tensor": "unsqueeze_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_80", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_103" + }, + { + "tensor": "mul_104" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_13", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_79" + }, + { + "tensor": "slice_71" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_14", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_80" + }, + { + "tensor": "slice_73" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_2", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_14" + }, + { + "tensor": "cat_14" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_3", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_15" + }, + { + "tensor": "transpose_38" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_17", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_2" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_4", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_17" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_11", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_4" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_4", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_11" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_18", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_3" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_5", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_18" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_12", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_5" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_5", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_12" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_39", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_4" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_3", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_13" + }, + { + "tensor": "transpose_39" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_105", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_3" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_81", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_105" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_1", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_81" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_4", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_1" + }, + { + "tensor": "_unsafe_view_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_40", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_4" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_13", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_40" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_51", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_13" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_7", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_106", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_51" + }, + { + "tensor": "sigmoid_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_58", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_106" + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_82", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_72" + }, + { + "tensor": "linear_58" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_26", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_82" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_25", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_26" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_83", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_25" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_37", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_83" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_107", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_82" + }, + { + "tensor": "rsqrt_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_84", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_108", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_107" + }, + { + "tensor": "add_84" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_19", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_108" + }, + { + "tensor": "add_82" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_59", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_19" + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_19", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_59" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_60", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_19" + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_109", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_19" + }, + { + "tensor": "linear_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_61", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_109" + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_85", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_82" + }, + { + "tensor": "linear_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_27", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_85" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_26", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_27" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_86", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_26" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_38", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_86" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_110", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_85" + }, + { + "tensor": "rsqrt_38" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_87", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_111", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_110" + }, + { + "tensor": "add_87" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_20", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_111" + }, + { + "tensor": "add_85" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_62", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_20" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_41", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_62" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_63", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_20" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_52", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_63" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_64", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_20" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_65", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_20" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_15", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_16" + }, + { + "tensor": "transpose_41" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_78", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_15" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_6", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_19", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_6" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_6", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_15" + }, + { + "tensor": "unsqueeze_19" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_79", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_6" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_20", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_79" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_80", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_20" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_42", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_80" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_81", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_42" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_82", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_42" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_83", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_42" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_53", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_81" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_54", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_82" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_55", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_83" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_8", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_64" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_6", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_10", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_88", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_65" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_6", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_88" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_112", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_10" + }, + { + "tensor": "softplus_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_113", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_53" + }, + { + "tensor": "view_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_13", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_113" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_89", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_13" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_39", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_89" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_114", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_53" + }, + { + "tensor": "rsqrt_39" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_115", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_54" + }, + { + "tensor": "view_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_14", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_115" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_90", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_14" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_40", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_90" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_116", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_54" + }, + { + "tensor": "rsqrt_40" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_6", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_114" + }, + { + "tensor": "mul_116" + }, + { + "tensor": "view_55" + }, + { + "tensor": "mul_112" + }, + { + "tensor": "sigmoid_8" + }, + { + "tensor": "states_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_84", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_6" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_85", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_6" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_43", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_85" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_14", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_43" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_56", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_14" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_57", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_52" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_28", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_56" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_27", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_28" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_91", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_27" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_41", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_91" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_117", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_56" + }, + { + "tensor": "rsqrt_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_118", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_norm_weight" + }, + { + "tensor": "mul_117" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_21", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_57" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_119", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_118" + }, + { + "tensor": "silu_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_58", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_119" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_66", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_58" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_92", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_85" + }, + { + "tensor": "linear_66" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_29", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_92" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_28", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_29" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_93", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_28" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_42", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_93" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_120", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_92" + }, + { + "tensor": "rsqrt_42" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_94", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_121", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_120" + }, + { + "tensor": "add_94" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_21", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_121" + }, + { + "tensor": "add_92" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_67", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_21" + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_22", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_68", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_21" + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_122", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_22" + }, + { + "tensor": "linear_68" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_69", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_122" + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_95", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_92" + }, + { + "tensor": "linear_69" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_30", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_95" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_29", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_30" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_96", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_29" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_43", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_96" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_123", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_95" + }, + { + "tensor": "rsqrt_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_97", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_124", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_123" + }, + { + "tensor": "add_97" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_22", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_124" + }, + { + "tensor": "add_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_70", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_22" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_44", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_70" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_71", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_22" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_59", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_71" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_72", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_22" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_73", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_22" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_16", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_18" + }, + { + "tensor": "transpose_44" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_86", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_16" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_7", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_20", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_7" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_7", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_16" + }, + { + "tensor": "unsqueeze_20" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_87", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_7" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_23", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_87" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_88", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_23" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_45", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_88" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_89", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_45" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_90", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_45" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_91", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_45" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_60", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_89" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_61", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_90" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_62", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_91" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_9", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_72" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_7", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_11", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_98", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_73" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_7", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_98" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_125", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_11" + }, + { + "tensor": "softplus_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_126", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_60" + }, + { + "tensor": "view_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_15", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_126" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_99", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_15" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_44", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_99" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_127", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_60" + }, + { + "tensor": "rsqrt_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_128", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_61" + }, + { + "tensor": "view_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_16", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_128" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_100", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_16" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_45", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_100" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_129", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_61" + }, + { + "tensor": "rsqrt_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_7", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_127" + }, + { + "tensor": "mul_129" + }, + { + "tensor": "view_62" + }, + { + "tensor": "mul_125" + }, + { + "tensor": "sigmoid_9" + }, + { + "tensor": "states_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_92", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_7" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_93", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_7" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_46", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_93" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_15", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_46" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_63", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_15" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_64", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_59" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_31", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_63" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_30", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_31" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_101", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_30" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_46", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_101" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_130", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_63" + }, + { + "tensor": "rsqrt_46" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_131", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_norm_weight" + }, + { + "tensor": "mul_130" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_24", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_64" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_132", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_131" + }, + { + "tensor": "silu_24" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_65", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_132" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_74", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_65" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_102", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_95" + }, + { + "tensor": "linear_74" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_32", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_102" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_31", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_32" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_103", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_31" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_47", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_103" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_133", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_102" + }, + { + "tensor": "rsqrt_47" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_104", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_134", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_133" + }, + { + "tensor": "add_104" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_23", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_134" + }, + { + "tensor": "add_102" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_75", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_23" + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_25", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_76", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_23" + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_135", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_25" + }, + { + "tensor": "linear_76" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_77", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_135" + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_105", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_102" + }, + { + "tensor": "linear_77" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_33", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_105" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_32", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_33" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_106", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_32" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_48", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_106" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_136", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_105" + }, + { + "tensor": "rsqrt_48" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_107", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_137", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_136" + }, + { + "tensor": "add_107" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_24", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_137" + }, + { + "tensor": "add_105" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_78", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_24" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_47", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_78" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_79", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_24" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_66", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_79" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_80", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_24" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_81", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_24" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_17", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_20" + }, + { + "tensor": "transpose_47" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_94", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_17" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_8", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_21", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_8" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_8", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_17" + }, + { + "tensor": "unsqueeze_21" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_95", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_8" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_26", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_96", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_26" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_48", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_96" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_97", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_48" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_98", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_48" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_99", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_48" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_67", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_97" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_68", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_98" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_69", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_99" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_10", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_80" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_8", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_12", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_108", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_81" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_8", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_108" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_138", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_12" + }, + { + "tensor": "softplus_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_139", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_67" + }, + { + "tensor": "view_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_17", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_139" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_109", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_17" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_49", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_109" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_140", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_67" + }, + { + "tensor": "rsqrt_49" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_141", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_68" + }, + { + "tensor": "view_68" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_18", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_141" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_110", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_18" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_50", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_110" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_142", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_68" + }, + { + "tensor": "rsqrt_50" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_8", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_140" + }, + { + "tensor": "mul_142" + }, + { + "tensor": "view_69" + }, + { + "tensor": "mul_138" + }, + { + "tensor": "sigmoid_10" + }, + { + "tensor": "states_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_100", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_8" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_101", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_8" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_49", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_101" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_16", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_49" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_70", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_16" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_71", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_66" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_34", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_70" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_33", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_34" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_111", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_33" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_51", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_111" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_143", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_70" + }, + { + "tensor": "rsqrt_51" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_144", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_norm_weight" + }, + { + "tensor": "mul_143" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_27", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_71" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_145", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_144" + }, + { + "tensor": "silu_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_72", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_145" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_82", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_72" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_112", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_105" + }, + { + "tensor": "linear_82" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_35", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_112" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_34", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_35" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_113", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_34" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_52", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_146", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_112" + }, + { + "tensor": "rsqrt_52" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_114", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_147", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_146" + }, + { + "tensor": "add_114" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_25", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_147" + }, + { + "tensor": "add_112" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_83", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_25" + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_28", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_83" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_84", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_25" + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_148", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_28" + }, + { + "tensor": "linear_84" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_85", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_148" + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_115", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_112" + }, + { + "tensor": "linear_85" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_36", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_115" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_35", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_36" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_116", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_35" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_53", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_116" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_149", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_115" + }, + { + "tensor": "rsqrt_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_117", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_150", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_149" + }, + { + "tensor": "add_117" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_26", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_150" + }, + { + "tensor": "add_115" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_86", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_26" + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_73", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_86" + }, + [ + 1, + 16, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_102", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_73" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_103", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_73" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_17", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_103" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_6", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_17" + }, + [ + 1, + 16, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_74", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_102" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_37", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_74" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_36", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_37" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_118", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_36" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_54", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_118" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_151", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_74" + }, + { + "tensor": "rsqrt_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_119", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_152", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_151" + }, + { + "tensor": "add_119" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_27", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_152" + }, + { + "tensor": "view_74" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_50", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_27" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_87", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_26" + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_75", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_87" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_38", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_75" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_37", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_38" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_120", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_37" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_55", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_120" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_153", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_75" + }, + { + "tensor": "rsqrt_55" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_121", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_154", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_153" + }, + { + "tensor": "add_121" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_28", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_154" + }, + { + "tensor": "view_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_51", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_28" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_88", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_26" + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_76", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_88" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_52", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_76" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_22", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_23", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_104", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_50" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_105", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_50" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_106", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_51" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_107", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_51" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_155", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_104" + }, + { + "tensor": "unsqueeze_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_108", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_104" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_109", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_104" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_13", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_109" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_18", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_13" + }, + { + "tensor": "slice_108" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_156", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_18" + }, + { + "tensor": "unsqueeze_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_122", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_155" + }, + { + "tensor": "mul_156" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_157", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_106" + }, + { + "tensor": "unsqueeze_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_110", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_106" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_111", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_106" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_14", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_111" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_19", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_14" + }, + { + "tensor": "slice_110" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_158", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_19" + }, + { + "tensor": "unsqueeze_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_123", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_157" + }, + { + "tensor": "mul_158" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_20", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_122" + }, + { + "tensor": "slice_105" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_21", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_123" + }, + { + "tensor": "slice_107" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_4", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_22" + }, + { + "tensor": "cat_21" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_5", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_23" + }, + { + "tensor": "transpose_52" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_24", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_4" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_6", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_24" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_18", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_6" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_7", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_18" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_25", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_5" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_7", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_25" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_19", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_7" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_8", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_19" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_53", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_7" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_5", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_20" + }, + { + "tensor": "transpose_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_159", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_5" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_124", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_159" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_2", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_124" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_6", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_2" + }, + { + "tensor": "_unsafe_view_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_54", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_6" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_20", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_54" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_77", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_20" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_11", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_160", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_77" + }, + { + "tensor": "sigmoid_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_89", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_160" + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_125", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_115" + }, + { + "tensor": "linear_89" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_39", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_125" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_38", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_39" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_126", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_38" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_56", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_126" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_161", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_125" + }, + { + "tensor": "rsqrt_56" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_127", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_162", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_161" + }, + { + "tensor": "add_127" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_29", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_162" + }, + { + "tensor": "add_125" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_90", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_29" + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_29", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_90" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_91", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_29" + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_163", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_29" + }, + { + "tensor": "linear_91" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_92", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_163" + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_128", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_125" + }, + { + "tensor": "linear_92" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_40", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_128" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_39", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_40" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_129", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_39" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_57", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_129" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_164", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_128" + }, + { + "tensor": "rsqrt_57" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_130", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_165", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_164" + }, + { + "tensor": "add_130" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_30", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_165" + }, + { + "tensor": "add_128" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_93", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_30" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_55", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_93" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_94", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_30" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_78", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_94" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_95", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_30" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_96", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_30" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_22", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_24" + }, + { + "tensor": "transpose_55" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_112", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_22" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_9", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_26", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_9" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_9", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_22" + }, + { + "tensor": "unsqueeze_26" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_113", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_9" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_30", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_114", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_30" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_56", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_114" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_115", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_56" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_116", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_56" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_117", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_56" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_79", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_115" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_80", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_116" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_81", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_117" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_12", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_9", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_15", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_131", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_96" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_9", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_131" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_166", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_15" + }, + { + "tensor": "softplus_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_167", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_79" + }, + { + "tensor": "view_79" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_19", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_167" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_132", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_19" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_58", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_132" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_168", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_79" + }, + { + "tensor": "rsqrt_58" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_169", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_80" + }, + { + "tensor": "view_80" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_20", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_169" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_133", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_20" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_59", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_133" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_170", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_80" + }, + { + "tensor": "rsqrt_59" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_9", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_168" + }, + { + "tensor": "mul_170" + }, + { + "tensor": "view_81" + }, + { + "tensor": "mul_166" + }, + { + "tensor": "sigmoid_12" + }, + { + "tensor": "states_25" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_118", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_9" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_119", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_9" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_57", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_119" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_21", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_57" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_82", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_21" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_83", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_78" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_41", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_82" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_40", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_41" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_134", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_40" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_60", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_134" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_171", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_82" + }, + { + "tensor": "rsqrt_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_172", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_norm_weight" + }, + { + "tensor": "mul_171" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_31", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_83" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_173", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_172" + }, + { + "tensor": "silu_31" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_84", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_173" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_97", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_84" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_135", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_128" + }, + { + "tensor": "linear_97" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_42", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_135" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_41", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_42" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_136", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_41" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_61", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_136" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_174", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_135" + }, + { + "tensor": "rsqrt_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_137", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_175", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_174" + }, + { + "tensor": "add_137" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_31", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_175" + }, + { + "tensor": "add_135" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_98", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_31" + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_32", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_98" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_99", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_31" + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_176", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_32" + }, + { + "tensor": "linear_99" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_100", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_176" + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_138", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_135" + }, + { + "tensor": "linear_100" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_43", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_138" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_42", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_43" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_139", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_42" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_62", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_139" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_177", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_138" + }, + { + "tensor": "rsqrt_62" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_140", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_178", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_177" + }, + { + "tensor": "add_140" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_32", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_178" + }, + { + "tensor": "add_138" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_101", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_32" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_58", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_101" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_102", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_32" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_85", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_102" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_103", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_32" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_104", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_32" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_23", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_26" + }, + { + "tensor": "transpose_58" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_120", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_23" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_10", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_27", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_10" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_10", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_23" + }, + { + "tensor": "unsqueeze_27" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_121", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_10" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_33", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_121" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_122", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_33" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_59", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_122" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_123", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_59" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_124", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_59" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_125", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_59" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_86", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_123" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_87", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_124" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_88", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_125" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_13", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_103" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_10", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_16", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_141", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_104" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_10", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_141" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_179", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_16" + }, + { + "tensor": "softplus_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_180", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_86" + }, + { + "tensor": "view_86" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_21", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_180" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_142", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_21" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_63", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_142" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_181", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_86" + }, + { + "tensor": "rsqrt_63" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_182", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_87" + }, + { + "tensor": "view_87" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_22", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_182" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_143", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_22" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_64", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_143" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_183", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_87" + }, + { + "tensor": "rsqrt_64" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_10", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_181" + }, + { + "tensor": "mul_183" + }, + { + "tensor": "view_88" + }, + { + "tensor": "mul_179" + }, + { + "tensor": "sigmoid_13" + }, + { + "tensor": "states_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_126", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_10" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_127", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_10" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_60", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_127" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_22", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_60" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_89", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_22" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_90", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_85" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_44", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_89" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_43", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_44" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_144", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_43" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_65", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_144" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_184", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_89" + }, + { + "tensor": "rsqrt_65" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_185", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_norm_weight" + }, + { + "tensor": "mul_184" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_34", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_90" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_186", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_185" + }, + { + "tensor": "silu_34" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_91", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_186" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_105", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_91" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_145", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_138" + }, + { + "tensor": "linear_105" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_45", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_145" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_44", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_45" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_146", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_44" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_66", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_146" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_187", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_145" + }, + { + "tensor": "rsqrt_66" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_147", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_188", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_187" + }, + { + "tensor": "add_147" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_33", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_188" + }, + { + "tensor": "add_145" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_106", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_33" + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_35", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_106" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_107", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_33" + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_189", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_35" + }, + { + "tensor": "linear_107" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_108", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_189" + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_148", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_145" + }, + { + "tensor": "linear_108" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_46", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_148" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_45", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_46" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_149", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_45" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_67", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_149" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_190", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_148" + }, + { + "tensor": "rsqrt_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_150", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_191", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_190" + }, + { + "tensor": "add_150" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_34", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_191" + }, + { + "tensor": "add_148" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_109", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_34" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_61", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_109" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_110", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_34" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_92", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_110" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_111", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_34" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_112", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_34" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_24", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_28" + }, + { + "tensor": "transpose_61" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_128", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_24" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_11", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_28", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_11" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_11", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_24" + }, + { + "tensor": "unsqueeze_28" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_129", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_11" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_36", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_129" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_130", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_36" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_62", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_130" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_131", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_62" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_132", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_62" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_133", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_62" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_93", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_131" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_94", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_132" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_95", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_133" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_14", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_111" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_11", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_17", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_151", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_112" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_11", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_151" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_192", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_17" + }, + { + "tensor": "softplus_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_193", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_93" + }, + { + "tensor": "view_93" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_23", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_193" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_152", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_23" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_68", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_152" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_194", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_93" + }, + { + "tensor": "rsqrt_68" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_195", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_94" + }, + { + "tensor": "view_94" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_24", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_195" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_153", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_24" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_69", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_153" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_196", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_94" + }, + { + "tensor": "rsqrt_69" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_11", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_194" + }, + { + "tensor": "mul_196" + }, + { + "tensor": "view_95" + }, + { + "tensor": "mul_192" + }, + { + "tensor": "sigmoid_14" + }, + { + "tensor": "states_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_134", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_11" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_135", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_11" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_63", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_135" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_23", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_63" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_96", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_23" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_97", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_92" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_47", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_96" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_46", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_47" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_154", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_46" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_70", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_154" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_197", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_96" + }, + { + "tensor": "rsqrt_70" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_198", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_norm_weight" + }, + { + "tensor": "mul_197" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_37", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_97" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_199", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_198" + }, + { + "tensor": "silu_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_98", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_199" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_113", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_98" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_155", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_148" + }, + { + "tensor": "linear_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_48", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_155" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_47", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_48" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_156", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_47" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_71", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_156" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_200", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_155" + }, + { + "tensor": "rsqrt_71" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_157", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_201", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_200" + }, + { + "tensor": "add_157" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_35", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_201" + }, + { + "tensor": "add_155" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_114", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_35" + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_38", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_114" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_115", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_35" + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_202", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_38" + }, + { + "tensor": "linear_115" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_116", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_202" + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_158", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_155" + }, + { + "tensor": "linear_116" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_49", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_158" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_48", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_49" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_159", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_48" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_72", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_159" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_203", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_158" + }, + { + "tensor": "rsqrt_72" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_160", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_204", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_203" + }, + { + "tensor": "add_160" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_36", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_204" + }, + { + "tensor": "add_158" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_117", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_36" + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_99", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_117" + }, + [ + 1, + 16, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_136", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_99" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_137", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_99" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_24", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_137" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_9", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_24" + }, + [ + 1, + 16, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_100", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_136" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_50", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_100" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_49", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_50" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_161", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_49" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_73", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_161" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_205", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_100" + }, + { + "tensor": "rsqrt_73" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_162", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_206", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_205" + }, + { + "tensor": "add_162" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_37", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_206" + }, + { + "tensor": "view_100" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_64", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_37" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_118", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_36" + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_101", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_118" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_51", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_101" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_50", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_51" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_163", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_50" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_74", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_163" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_207", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_101" + }, + { + "tensor": "rsqrt_74" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_164", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_208", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_207" + }, + { + "tensor": "add_164" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_38", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_208" + }, + { + "tensor": "view_101" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_65", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_38" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_119", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_36" + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_102", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_119" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_66", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_102" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_29", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_30", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_138", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_64" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_139", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_64" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_140", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_65" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_141", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_65" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_209", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_138" + }, + { + "tensor": "unsqueeze_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_142", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_138" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_143", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_138" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_18", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_143" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_25", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_18" + }, + { + "tensor": "slice_142" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_210", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_25" + }, + { + "tensor": "unsqueeze_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_165", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_209" + }, + { + "tensor": "mul_210" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_211", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_140" + }, + { + "tensor": "unsqueeze_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_144", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_140" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_145", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_140" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_19", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_145" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_26", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_19" + }, + { + "tensor": "slice_144" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_212", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_26" + }, + { + "tensor": "unsqueeze_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_166", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_211" + }, + { + "tensor": "mul_212" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_27", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_165" + }, + { + "tensor": "slice_139" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_28", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_166" + }, + { + "tensor": "slice_141" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_6", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_30" + }, + { + "tensor": "cat_28" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_7", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_31" + }, + { + "tensor": "transpose_66" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_31", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_6" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_8", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_31" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_25", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_8" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_10", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_25" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_32", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_7" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_9", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_32" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_26", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_9" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_11", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_26" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_67", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_10" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_7", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_27" + }, + { + "tensor": "transpose_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_213", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_7" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_167", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_213" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_3", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_167" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_8", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_3" + }, + { + "tensor": "_unsafe_view_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_68", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_8" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_27", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_68" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_103", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_27" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_15", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_214", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_103" + }, + { + "tensor": "sigmoid_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_120", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_214" + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_168", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_158" + }, + { + "tensor": "linear_120" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_52", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_168" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_51", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_52" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_169", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_51" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_75", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_169" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_215", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_168" + }, + { + "tensor": "rsqrt_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_170", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_216", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_215" + }, + { + "tensor": "add_170" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_39", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_216" + }, + { + "tensor": "add_168" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_121", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_39" + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_39", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_121" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_122", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_39" + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_217", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_39" + }, + { + "tensor": "linear_122" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_123", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_217" + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_171", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_168" + }, + { + "tensor": "linear_123" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_53", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_171" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_52", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_53" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_172", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_52" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_76", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_172" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_218", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_171" + }, + { + "tensor": "rsqrt_76" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_173", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_219", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_218" + }, + { + "tensor": "add_173" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_40", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_219" + }, + { + "tensor": "add_171" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_124", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_40" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_69", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_124" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_125", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_40" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_104", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_125" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_126", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_40" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_127", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_40" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_29", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_32" + }, + { + "tensor": "transpose_69" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_146", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_29" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_12", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_33", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_12" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_12", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_29" + }, + { + "tensor": "unsqueeze_33" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_147", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_12" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_40", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_147" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_148", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_40" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_70", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_148" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_149", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_70" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_150", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_70" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_151", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_70" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_105", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_149" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_106", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_150" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_107", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_151" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_16", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_126" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_12", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_20", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_174", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_127" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_12", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_174" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_220", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_20" + }, + { + "tensor": "softplus_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_221", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_105" + }, + { + "tensor": "view_105" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_25", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_221" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_175", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_25" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_77", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_175" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_222", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_105" + }, + { + "tensor": "rsqrt_77" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_223", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_106" + }, + { + "tensor": "view_106" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_26", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_223" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_176", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_26" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_78", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_176" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_224", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_106" + }, + { + "tensor": "rsqrt_78" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_12", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_222" + }, + { + "tensor": "mul_224" + }, + { + "tensor": "view_107" + }, + { + "tensor": "mul_220" + }, + { + "tensor": "sigmoid_16" + }, + { + "tensor": "states_33" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_152", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_12" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_153", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_12" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_71", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_153" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_28", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_71" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_108", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_28" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_109", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_104" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_54", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_108" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_53", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_54" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_177", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_53" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_79", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_177" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_225", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_108" + }, + { + "tensor": "rsqrt_79" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_226", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_norm_weight" + }, + { + "tensor": "mul_225" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_41", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_109" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_227", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_226" + }, + { + "tensor": "silu_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_110", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_227" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_128", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_110" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_178", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_171" + }, + { + "tensor": "linear_128" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_55", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_178" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_54", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_55" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_179", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_54" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_80", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_179" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_228", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_178" + }, + { + "tensor": "rsqrt_80" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_180", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_229", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_228" + }, + { + "tensor": "add_180" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_41", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_229" + }, + { + "tensor": "add_178" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_129", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_41" + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_42", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_129" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_130", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_41" + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_230", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_42" + }, + { + "tensor": "linear_130" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_131", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_230" + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_181", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_178" + }, + { + "tensor": "linear_131" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_56", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_181" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_55", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_56" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_182", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_55" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_81", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_182" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_231", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_181" + }, + { + "tensor": "rsqrt_81" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_183", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_232", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_231" + }, + { + "tensor": "add_183" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_42", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_232" + }, + { + "tensor": "add_181" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_132", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_42" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_72", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_132" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_133", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_42" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_111", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_133" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_134", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_42" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_135", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_42" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_30", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_34" + }, + { + "tensor": "transpose_72" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_154", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_30" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_13", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_34", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_13" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_13", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_30" + }, + { + "tensor": "unsqueeze_34" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_155", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_13" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_43", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_155" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_156", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_43" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_73", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_156" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_157", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_73" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_158", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_73" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_159", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_73" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_112", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_157" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_113", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_158" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_114", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_159" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_17", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_134" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_13", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_21", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_184", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_135" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_13", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_184" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_233", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_21" + }, + { + "tensor": "softplus_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_234", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_112" + }, + { + "tensor": "view_112" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_27", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_234" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_185", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_27" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_82", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_185" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_235", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_112" + }, + { + "tensor": "rsqrt_82" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_236", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_113" + }, + { + "tensor": "view_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_28", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_236" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_186", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_28" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_83", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_186" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_237", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_113" + }, + { + "tensor": "rsqrt_83" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_13", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_235" + }, + { + "tensor": "mul_237" + }, + { + "tensor": "view_114" + }, + { + "tensor": "mul_233" + }, + { + "tensor": "sigmoid_17" + }, + { + "tensor": "states_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_160", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_13" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_161", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_13" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_74", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_161" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_29", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_74" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_115", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_29" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_116", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_111" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_57", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_115" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_56", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_57" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_187", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_56" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_84", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_187" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_238", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_115" + }, + { + "tensor": "rsqrt_84" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_239", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_norm_weight" + }, + { + "tensor": "mul_238" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_44", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_116" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_240", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_239" + }, + { + "tensor": "silu_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_117", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_240" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_136", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_117" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_188", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_181" + }, + { + "tensor": "linear_136" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_58", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_188" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_57", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_58" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_189", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_57" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_85", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_189" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_241", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_188" + }, + { + "tensor": "rsqrt_85" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_190", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_242", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_241" + }, + { + "tensor": "add_190" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_43", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_242" + }, + { + "tensor": "add_188" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_137", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_43" + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_45", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_137" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_138", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_43" + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_243", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_45" + }, + { + "tensor": "linear_138" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_139", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_243" + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_191", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_188" + }, + { + "tensor": "linear_139" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_59", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_191" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_58", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_59" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_192", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_58" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_86", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_192" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_244", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_191" + }, + { + "tensor": "rsqrt_86" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_193", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_245", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_244" + }, + { + "tensor": "add_193" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_44", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_245" + }, + { + "tensor": "add_191" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_140", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_44" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_75", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_140" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_141", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_44" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_118", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_141" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_142", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_44" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_143", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_44" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_31", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_36" + }, + { + "tensor": "transpose_75" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_162", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_31" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_14", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_35", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_14" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_14", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_31" + }, + { + "tensor": "unsqueeze_35" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_163", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_14" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_46", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_163" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_164", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_46" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_76", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_164" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_165", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_76" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_166", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_76" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_167", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_76" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_119", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_165" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_120", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_166" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_121", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_167" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_18", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_142" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_14", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_22", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_194", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_143" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_14", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_194" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_246", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_22" + }, + { + "tensor": "softplus_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_247", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_119" + }, + { + "tensor": "view_119" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_29", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_247" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_195", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_29" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_87", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_195" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_248", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_119" + }, + { + "tensor": "rsqrt_87" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_249", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_120" + }, + { + "tensor": "view_120" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_30", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_249" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_196", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_30" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_88", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_196" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_250", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_120" + }, + { + "tensor": "rsqrt_88" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_14", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_248" + }, + { + "tensor": "mul_250" + }, + { + "tensor": "view_121" + }, + { + "tensor": "mul_246" + }, + { + "tensor": "sigmoid_18" + }, + { + "tensor": "states_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_168", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_14" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_169", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_14" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_77", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_169" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_30", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_77" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_122", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_30" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_123", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_118" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_60", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_122" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_59", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_60" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_197", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_59" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_89", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_197" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_251", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_122" + }, + { + "tensor": "rsqrt_89" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_252", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_norm_weight" + }, + { + "tensor": "mul_251" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_47", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_123" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_253", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_252" + }, + { + "tensor": "silu_47" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_124", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_253" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_144", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_124" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_198", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_191" + }, + { + "tensor": "linear_144" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_61", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_198" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_60", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_61" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_199", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_60" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_90", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_199" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_254", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_198" + }, + { + "tensor": "rsqrt_90" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_200", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_255", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_254" + }, + { + "tensor": "add_200" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_45", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_255" + }, + { + "tensor": "add_198" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_145", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_45" + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_48", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_145" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_146", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_45" + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_256", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_48" + }, + { + "tensor": "linear_146" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_147", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_256" + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_201", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_198" + }, + { + "tensor": "linear_147" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_62", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_201" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_61", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_62" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_202", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_61" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_91", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_202" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_257", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_201" + }, + { + "tensor": "rsqrt_91" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_203", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_258", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_257" + }, + { + "tensor": "add_203" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_46", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_258" + }, + { + "tensor": "add_201" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_148", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_46" + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_125", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_148" + }, + [ + 1, + 16, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_170", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_125" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_171", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_125" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_31", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_171" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_12", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_31" + }, + [ + 1, + 16, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_126", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_170" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_63", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_126" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_62", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_63" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_204", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_62" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_92", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_204" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_259", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_126" + }, + { + "tensor": "rsqrt_92" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_205", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_260", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_259" + }, + { + "tensor": "add_205" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_47", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_260" + }, + { + "tensor": "view_126" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_78", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_47" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_149", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_46" + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_127", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_149" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_64", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_127" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_63", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_64" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_206", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_63" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_93", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_206" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_261", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_127" + }, + { + "tensor": "rsqrt_93" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_207", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_262", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_261" + }, + { + "tensor": "add_207" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_48", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_262" + }, + { + "tensor": "view_127" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_79", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_48" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_150", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_46" + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_128", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_150" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_80", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_128" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_36", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_37", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_172", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_78" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_173", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_78" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_174", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_79" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_175", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_79" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_263", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_172" + }, + { + "tensor": "unsqueeze_36" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_176", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_172" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_177", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_172" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_23", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_177" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_32", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_23" + }, + { + "tensor": "slice_176" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_264", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_32" + }, + { + "tensor": "unsqueeze_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_208", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_263" + }, + { + "tensor": "mul_264" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_265", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_174" + }, + { + "tensor": "unsqueeze_36" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_178", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_174" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_179", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_174" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_24", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_179" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_33", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_24" + }, + { + "tensor": "slice_178" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_266", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_33" + }, + { + "tensor": "unsqueeze_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_209", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_265" + }, + { + "tensor": "mul_266" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_34", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_208" + }, + { + "tensor": "slice_173" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_35", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_209" + }, + { + "tensor": "slice_175" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_8", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_38" + }, + { + "tensor": "cat_35" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_9", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_39" + }, + { + "tensor": "transpose_80" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_38", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_8" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_10", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_38" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_32", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_10" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_13", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_32" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_39", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_9" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_11", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_39" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_33", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_11" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_14", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_33" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_81", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_13" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_9", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_34" + }, + { + "tensor": "transpose_81" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_267", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_9" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_210", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_267" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_4", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_210" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_10", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_4" + }, + { + "tensor": "_unsafe_view_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_82", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_10" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_34", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_82" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_129", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_34" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_19", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_268", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_129" + }, + { + "tensor": "sigmoid_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_151", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_268" + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_211", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_201" + }, + { + "tensor": "linear_151" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_65", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_211" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_64", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_65" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_212", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_64" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_94", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_212" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_269", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_211" + }, + { + "tensor": "rsqrt_94" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_213", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_270", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_269" + }, + { + "tensor": "add_213" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_49", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_270" + }, + { + "tensor": "add_211" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_152", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_49" + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_49", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_152" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_153", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_49" + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_271", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_49" + }, + { + "tensor": "linear_153" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_154", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_271" + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_214", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_211" + }, + { + "tensor": "linear_154" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_66", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_214" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_65", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_66" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_215", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_65" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_95", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_215" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_272", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_214" + }, + { + "tensor": "rsqrt_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_216", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_273", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_272" + }, + { + "tensor": "add_216" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_50", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_273" + }, + { + "tensor": "add_214" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_155", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_50" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_83", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_155" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_156", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_50" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_130", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_156" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_157", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_50" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_158", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_50" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_36", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_40" + }, + { + "tensor": "transpose_83" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_180", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_36" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_15", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_40", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_15" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_15", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_36" + }, + { + "tensor": "unsqueeze_40" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_181", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_15" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_50", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_181" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_182", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_50" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_84", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_182" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_183", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_84" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_184", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_84" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_185", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_84" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_131", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_183" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_132", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_184" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_133", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_185" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_20", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_157" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_15", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_25", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_217", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_158" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_15", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_217" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_274", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_25" + }, + { + "tensor": "softplus_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_275", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_131" + }, + { + "tensor": "view_131" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_31", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_275" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_218", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_31" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_96", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_218" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_276", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_131" + }, + { + "tensor": "rsqrt_96" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_277", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_132" + }, + { + "tensor": "view_132" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_32", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_277" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_219", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_32" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_97", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_219" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_278", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_132" + }, + { + "tensor": "rsqrt_97" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_15", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_276" + }, + { + "tensor": "mul_278" + }, + { + "tensor": "view_133" + }, + { + "tensor": "mul_274" + }, + { + "tensor": "sigmoid_20" + }, + { + "tensor": "states_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_186", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_15" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_187", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_15" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_85", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_187" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_35", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_85" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_134", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_35" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_135", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_130" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_67", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_134" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_66", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_67" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_220", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_66" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_98", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_220" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_279", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_134" + }, + { + "tensor": "rsqrt_98" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_280", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_norm_weight" + }, + { + "tensor": "mul_279" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_51", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_135" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_281", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_280" + }, + { + "tensor": "silu_51" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_136", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_281" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_159", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_136" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_221", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_214" + }, + { + "tensor": "linear_159" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_68", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_221" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_67", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_68" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_222", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_67" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_99", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_222" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_282", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_221" + }, + { + "tensor": "rsqrt_99" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_223", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_283", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_282" + }, + { + "tensor": "add_223" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_51", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_283" + }, + { + "tensor": "add_221" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_160", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_51" + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_52", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_160" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_161", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_51" + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_284", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_52" + }, + { + "tensor": "linear_161" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_162", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_284" + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_224", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_221" + }, + { + "tensor": "linear_162" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_69", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_224" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_68", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_69" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_225", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_68" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_100", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_225" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_285", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_224" + }, + { + "tensor": "rsqrt_100" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_226", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_286", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_285" + }, + { + "tensor": "add_226" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_52", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_286" + }, + { + "tensor": "add_224" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_163", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_52" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_86", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_163" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_164", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_52" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_137", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_164" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_165", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_52" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_166", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_52" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_37", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_42" + }, + { + "tensor": "transpose_86" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_188", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_37" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_16", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_41", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_16" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_16", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_37" + }, + { + "tensor": "unsqueeze_41" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_189", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_16" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_53", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_189" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_190", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_53" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_87", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_190" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_191", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_87" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_192", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_87" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_193", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_87" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_138", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_191" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_139", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_192" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_140", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_193" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_21", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_165" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_16", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_26", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_227", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_166" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_16", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_227" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_287", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_26" + }, + { + "tensor": "softplus_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_288", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_138" + }, + { + "tensor": "view_138" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_33", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_288" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_228", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_33" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_101", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_228" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_289", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_138" + }, + { + "tensor": "rsqrt_101" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_290", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_139" + }, + { + "tensor": "view_139" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_34", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_290" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_229", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_34" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_102", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_229" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_291", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_139" + }, + { + "tensor": "rsqrt_102" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_16", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_289" + }, + { + "tensor": "mul_291" + }, + { + "tensor": "view_140" + }, + { + "tensor": "mul_287" + }, + { + "tensor": "sigmoid_21" + }, + { + "tensor": "states_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_194", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_16" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_195", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_16" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_88", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_195" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_36", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_88" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_141", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_36" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_142", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_137" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_70", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_141" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_69", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_70" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_230", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_69" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_103", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_230" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_292", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_141" + }, + { + "tensor": "rsqrt_103" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_293", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_norm_weight" + }, + { + "tensor": "mul_292" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_54", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_142" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_294", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_293" + }, + { + "tensor": "silu_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_143", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_294" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_167", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_143" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_231", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_224" + }, + { + "tensor": "linear_167" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_71", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_231" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_70", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_71" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_232", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_70" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_104", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_232" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_295", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_231" + }, + { + "tensor": "rsqrt_104" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_233", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_296", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_295" + }, + { + "tensor": "add_233" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_53", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_296" + }, + { + "tensor": "add_231" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_168", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_53" + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_55", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_168" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_169", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_53" + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_297", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_55" + }, + { + "tensor": "linear_169" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_170", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_297" + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_234", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_231" + }, + { + "tensor": "linear_170" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_72", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_234" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_71", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_72" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_235", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_71" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_105", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_235" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_298", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_234" + }, + { + "tensor": "rsqrt_105" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_236", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_299", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_298" + }, + { + "tensor": "add_236" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_54", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_299" + }, + { + "tensor": "add_234" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_171", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_54" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_89", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_171" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_172", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_54" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_144", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_172" + }, + [ + 1, + 16, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_173", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_54" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_174", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_54" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_38", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_44" + }, + { + "tensor": "transpose_89" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_196", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_38" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_17", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_42", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_17" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_17", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_38" + }, + { + "tensor": "unsqueeze_42" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_197", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_17" + }, + 2, + 0, + 20 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_56", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_197" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_198", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_56" + }, + 2, + -16, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_90", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_198" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_199", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_90" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_200", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_90" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_201", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_90" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_145", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_199" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_146", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_200" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_147", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_201" + }, + [ + 1, + 16, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_22", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_173" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_17", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_27", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_237", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_174" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_17", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_237" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_300", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_27" + }, + { + "tensor": "softplus_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_301", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_145" + }, + { + "tensor": "view_145" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_35", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_301" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_238", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_35" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_106", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_238" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_302", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_145" + }, + { + "tensor": "rsqrt_106" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_303", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_146" + }, + { + "tensor": "view_146" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_36", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_303" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_239", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_36" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_107", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_239" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_304", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_146" + }, + { + "tensor": "rsqrt_107" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_17", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_302" + }, + { + "tensor": "mul_304" + }, + { + "tensor": "view_147" + }, + { + "tensor": "mul_300" + }, + { + "tensor": "sigmoid_22" + }, + { + "tensor": "states_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_202", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_17" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_203", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_17" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_91", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_203" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_37", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_91" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_148", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_37" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_149", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_144" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_73", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_148" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_72", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_73" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_240", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_72" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_108", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_240" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_305", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_148" + }, + { + "tensor": "rsqrt_108" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_306", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_norm_weight" + }, + { + "tensor": "mul_305" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_57", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_149" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_307", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_306" + }, + { + "tensor": "silu_57" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_150", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_307" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_175", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_150" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_241", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_234" + }, + { + "tensor": "linear_175" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_74", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_241" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_73", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_74" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_242", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_73" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_109", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_242" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_308", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_241" + }, + { + "tensor": "rsqrt_109" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_243", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_309", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_308" + }, + { + "tensor": "add_243" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_55", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_309" + }, + { + "tensor": "add_241" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_176", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_55" + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_58", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_176" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_177", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_55" + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_310", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_58" + }, + { + "tensor": "linear_177" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_178", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_310" + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_244", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_241" + }, + { + "tensor": "linear_178" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_75", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_244" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_74", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_75" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_245", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_74" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_110", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_245" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_311", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_244" + }, + { + "tensor": "rsqrt_110" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_246", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_312", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_311" + }, + { + "tensor": "add_246" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_56", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_312" + }, + { + "tensor": "add_244" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_179", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_56" + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_151", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_179" + }, + [ + 1, + 16, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_204", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_151" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_205", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_151" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_38", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_205" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_15", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_38" + }, + [ + 1, + 16, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_152", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_204" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_76", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_152" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_75", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_76" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_247", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_75" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_111", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_247" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_313", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_152" + }, + { + "tensor": "rsqrt_111" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_248", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_314", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_313" + }, + { + "tensor": "add_248" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_57", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_314" + }, + { + "tensor": "view_152" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_92", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_57" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_180", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_56" + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_153", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_180" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_77", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_153" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_76", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_77" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_249", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_76" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_112", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_249" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_315", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_153" + }, + { + "tensor": "rsqrt_112" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_250", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_316", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_315" + }, + { + "tensor": "add_250" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_58", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_316" + }, + { + "tensor": "view_153" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_93", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_58" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_181", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_56" + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_154", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_181" + }, + [ + 1, + 16, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_94", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_154" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_43", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_44", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_206", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_92" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_207", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_92" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_208", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_93" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_209", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_93" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_317", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_206" + }, + { + "tensor": "unsqueeze_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_210", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_206" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_211", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_206" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_28", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_211" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_39", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_28" + }, + { + "tensor": "slice_210" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_318", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_39" + }, + { + "tensor": "unsqueeze_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_251", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_317" + }, + { + "tensor": "mul_318" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_319", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_208" + }, + { + "tensor": "unsqueeze_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_212", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_208" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_213", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_208" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_29", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_213" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_40", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_29" + }, + { + "tensor": "slice_212" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_320", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_40" + }, + { + "tensor": "unsqueeze_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_252", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_319" + }, + { + "tensor": "mul_320" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_41", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_251" + }, + { + "tensor": "slice_207" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_42", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_252" + }, + { + "tensor": "slice_209" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_10", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_46" + }, + { + "tensor": "cat_42" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_11", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_47" + }, + { + "tensor": "transpose_94" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_45", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_10" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_12", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_45" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_39", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_12" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_16", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_39" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_46", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_11" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_13", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_46" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_40", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_13" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_17", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_40" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_95", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_16" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_11", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_41" + }, + { + "tensor": "transpose_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_321", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_11" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_253", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_321" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_5", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_253" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_12", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_5" + }, + { + "tensor": "_unsafe_view_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_96", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_12" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_41", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_96" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_155", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_41" + }, + [ + 1, + 16, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_23", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_322", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_155" + }, + { + "tensor": "sigmoid_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_182", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_322" + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_254", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_244" + }, + { + "tensor": "linear_182" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_78", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_254" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_77", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_78" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_255", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_77" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_113", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_255" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_323", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_254" + }, + { + "tensor": "rsqrt_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_256", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_324", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_323" + }, + { + "tensor": "add_256" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_59", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_324" + }, + { + "tensor": "add_254" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_183", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_59" + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_59", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_183" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_184", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_59" + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_325", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_59" + }, + { + "tensor": "linear_184" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_185", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_325" + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_257", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_254" + }, + { + "tensor": "linear_185" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_79", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_257" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_78", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_79" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_258", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_78" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_114", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_258" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_326", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_257" + }, + { + "tensor": "rsqrt_114" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_259", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_327", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_326" + }, + { + "tensor": "add_259" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_60", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_327" + }, + { + "tensor": "add_257" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "alias", + "op": "aten.alias.default", + "args": [ + { + "tensor": "type_as_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1787, in forward\n logits = self.lm_head(hidden_states[:, slice_indices, :])\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_186", + "op": "aten.linear.default", + "args": [ + { + "tensor": "alias" + }, + { + "tensor": "p_model_model_language_model_embed_tokens_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1787, in forward\n logits = self.lm_head(hidden_states[:, slice_indices, :])\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + } + ], + "quantization": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007083918899297714, + "meanSquaredWeightError": 4.075269544046023e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005190409719944, + "meanSquaredWeightError": 1.0724062526890066e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014917552471160889, + "meanSquaredWeightError": 8.557733188752081e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001733013428747654, + "meanSquaredWeightError": 9.211684748322568e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001013568602502346, + "meanSquaredWeightError": 4.7989616724919415e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0420619398355484, + "meanSquaredWeightError": 9.975725561162108e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03571777045726776, + "meanSquaredWeightError": 1.8712555629463168e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.015869140625, + "meanSquaredWeightError": 9.336113180324901e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006478391587734222, + "meanSquaredWeightError": 1.841971197791281e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006430316716432571, + "meanSquaredWeightError": 1.099748647703791e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009914608672261238, + "meanSquaredWeightError": 8.36655100755479e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006439946591854095, + "meanSquaredWeightError": 5.678034487033301e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010154899209737778, + "meanSquaredWeightError": 4.161362809895763e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04477539658546448, + "meanSquaredWeightError": 9.298059922002722e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02817993052303791, + "meanSquaredWeightError": 1.585794393577089e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.017333984375, + "meanSquaredWeightError": 9.087384000849852e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03647809475660324, + "meanSquaredWeightError": 9.2817742824991e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03266602009534836, + "meanSquaredWeightError": 1.564963326927682e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0213623046875, + "meanSquaredWeightError": 9.297695555687824e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00043938105227425694, + "meanSquaredWeightError": 4.2963801405448976e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0004978924989700317, + "meanSquaredWeightError": 4.300909850485368e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007458776235580444, + "meanSquaredWeightError": 6.14908657325941e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005382634699344635, + "meanSquaredWeightError": 6.678897435108411e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006161201745271683, + "meanSquaredWeightError": 2.1588792975535398e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0004498325288295746, + "meanSquaredWeightError": 7.733858886638245e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010226964950561523, + "meanSquaredWeightError": 1.0263062399928913e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007535666227340698, + "meanSquaredWeightError": 6.758249515570469e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007804818451404572, + "meanSquaredWeightError": 3.901024836494571e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02197265625, + "meanSquaredWeightError": 8.842465035741043e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03532366454601288, + "meanSquaredWeightError": 1.9045186263610958e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01679687388241291, + "meanSquaredWeightError": 8.847858339322556e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006036236882209778, + "meanSquaredWeightError": 1.5833309774393456e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00047482550144195557, + "meanSquaredWeightError": 7.293367687566388e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009035095572471619, + "meanSquaredWeightError": 9.150482149777872e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007492424920201302, + "meanSquaredWeightError": 6.225117310521e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0012094092089682817, + "meanSquaredWeightError": 4.19806900353592e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0224609375, + "meanSquaredWeightError": 9.161611842500861e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04632568359375, + "meanSquaredWeightError": 2.056171069853008e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0159912109375, + "meanSquaredWeightError": 9.076159699361597e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007881708443164825, + "meanSquaredWeightError": 1.647453373720964e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005805529654026031, + "meanSquaredWeightError": 7.469957985506426e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0016993731260299683, + "meanSquaredWeightError": 9.388108956898122e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007583741098642349, + "meanSquaredWeightError": 7.161021109425292e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006665807450190187, + "meanSquaredWeightError": 4.480360082936841e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.037353515625, + "meanSquaredWeightError": 1.04351170193695e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.05363769829273224, + "meanSquaredWeightError": 2.7039927772420924e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01563110202550888, + "meanSquaredWeightError": 9.754314760357374e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.047607421875, + "meanSquaredWeightError": 1.168099856840854e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.05196359008550644, + "meanSquaredWeightError": 2.913413027272327e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0177001953125, + "meanSquaredWeightError": 1.0753815331554506e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00047194119542837143, + "meanSquaredWeightError": 4.19777368421137e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006257314234972, + "meanSquaredWeightError": 6.4926997112024765e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000988095998764038, + "meanSquaredWeightError": 7.445749794499079e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0011749267578125, + "meanSquaredWeightError": 1.0600491151535607e-08, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005680583417415619, + "meanSquaredWeightError": 1.9959742303399253e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005613304674625397, + "meanSquaredWeightError": 9.642544540611198e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0013072118163108826, + "meanSquaredWeightError": 1.371929148064055e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000817006453871727, + "meanSquaredWeightError": 7.562183768072828e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010419189929962158, + "meanSquaredWeightError": 4.178038359725633e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0301513671875, + "meanSquaredWeightError": 1.0534593002375914e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03706229105591774, + "meanSquaredWeightError": 3.2628909139020834e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01273193396627903, + "meanSquaredWeightError": 1.0341716460970929e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008150823414325714, + "meanSquaredWeightError": 2.1580238040996846e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00043829530477523804, + "meanSquaredWeightError": 7.784422884071773e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014495849609375, + "meanSquaredWeightError": 1.5511064432871535e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000626690685749054, + "meanSquaredWeightError": 8.258425054918916e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006689839065074921, + "meanSquaredWeightError": 4.077417070647016e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0342668816447258, + "meanSquaredWeightError": 1.0280851938659907e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.046142578125, + "meanSquaredWeightError": 3.7489039641513955e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0156511589884758, + "meanSquaredWeightError": 1.021842535919859e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007631815969944, + "meanSquaredWeightError": 1.8714940708264294e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005142334848642349, + "meanSquaredWeightError": 8.547659469115843e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014417767524719238, + "meanSquaredWeightError": 1.6978646044663037e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000848725438117981, + "meanSquaredWeightError": 8.861574585239396e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009453240782022476, + "meanSquaredWeightError": 3.604574860816001e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0531267449259758, + "meanSquaredWeightError": 1.0224184734397568e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04545222967863083, + "meanSquaredWeightError": 4.110375357413432e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.014916989952325821, + "meanSquaredWeightError": 1.0372635870226077e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04470040649175644, + "meanSquaredWeightError": 1.1175421832376742e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04736328125, + "meanSquaredWeightError": 3.7854479160159826e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01680908165872097, + "meanSquaredWeightError": 1.0865893500522361e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006805174052715302, + "meanSquaredWeightError": 4.193761338200375e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008737146854400635, + "meanSquaredWeightError": 9.042077309118213e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007929764688014984, + "meanSquaredWeightError": 7.600540641305997e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014283191412687302, + "meanSquaredWeightError": 1.78832415542729e-08, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000786246731877327, + "meanSquaredWeightError": 2.793629150232846e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008093155920505524, + "meanSquaredWeightError": 1.2136558424913346e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0012167382519692183, + "meanSquaredWeightError": 8.249294580764399e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007112771272659302, + "meanSquaredWeightError": 1.0718528287156914e-08, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0012726113200187683, + "meanSquaredWeightError": 4.84854734139617e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04365234822034836, + "meanSquaredWeightError": 9.266742040381359e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.05205077677965164, + "meanSquaredWeightError": 2.190386794609367e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02226562425494194, + "meanSquaredWeightError": 9.115561283579154e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007804818451404572, + "meanSquaredWeightError": 2.823620093295176e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0004916451871395111, + "meanSquaredWeightError": 1.1404850397411792e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0022107213735580444, + "meanSquaredWeightError": 1.7722628697924847e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001212533563375473, + "meanSquaredWeightError": 9.309404802593235e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001015007495880127, + "meanSquaredWeightError": 3.86098308879923e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03144530951976776, + "meanSquaredWeightError": 1.1349436590535333e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.043701171875, + "meanSquaredWeightError": 3.7038337268313626e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01273193396627903, + "meanSquaredWeightError": 1.1466236173873767e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000811239704489708, + "meanSquaredWeightError": 2.8647622940525252e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0004772273823618889, + "meanSquaredWeightError": 9.231112763075089e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001953125, + "meanSquaredWeightError": 1.5947463793963834e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010938271880149841, + "meanSquaredWeightError": 9.523899890950815e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010099634528160095, + "meanSquaredWeightError": 4.4895744899520196e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02999267354607582, + "meanSquaredWeightError": 1.2448520010366337e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03925780951976776, + "meanSquaredWeightError": 3.853431735478807e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02015380933880806, + "meanSquaredWeightError": 1.2184170827822527e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008477643132209778, + "meanSquaredWeightError": 3.4146133032209036e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007651001214981079, + "meanSquaredWeightError": 1.287605755351251e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0015244372189044952, + "meanSquaredWeightError": 1.2338504440378983e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001476377248764038, + "meanSquaredWeightError": 9.998956329582143e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001247616484761238, + "meanSquaredWeightError": 5.051828733115826e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0316162109375, + "meanSquaredWeightError": 1.3124330280334107e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.05973249301314354, + "meanSquaredWeightError": 4.17917181039229e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02794799953699112, + "meanSquaredWeightError": 1.3120700259605655e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0629359632730484, + "meanSquaredWeightError": 1.335452111561608e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0677926167845726, + "meanSquaredWeightError": 4.988436103303684e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0322265625, + "meanSquaredWeightError": 1.6804390270408476e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000582478940486908, + "meanSquaredWeightError": 3.9153045250373e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010428838431835175, + "meanSquaredWeightError": 8.011205920865905e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001420147716999054, + "meanSquaredWeightError": 8.69789218427286e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.002052607014775276, + "meanSquaredWeightError": 1.3341496796215324e-08, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007699094712734222, + "meanSquaredWeightError": 2.5097170563981308e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006843628361821175, + "meanSquaredWeightError": 1.136356786446413e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009746402502059937, + "meanSquaredWeightError": 8.813485941061572e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000755971297621727, + "meanSquaredWeightError": 1.1121874976538493e-08, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0011303499341011047, + "meanSquaredWeightError": 4.848550450020639e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03278808295726776, + "meanSquaredWeightError": 8.690146273693244e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0361328125, + "meanSquaredWeightError": 3.182140517310472e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01421508751809597, + "meanSquaredWeightError": 8.130265882755339e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03572387620806694, + "meanSquaredWeightError": 8.140267482303898e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02766112983226776, + "meanSquaredWeightError": 3.193840257154079e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01599992997944355, + "meanSquaredWeightError": 7.776296229167201e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00040658097714185715, + "meanSquaredWeightError": 4.779415085920391e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0012706853449344635, + "meanSquaredWeightError": 4.892196425743123e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009188912808895111, + "meanSquaredWeightError": 1.1496103624608622e-08, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005174772813916206, + "meanSquaredWeightError": 5.1614277296607725e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006728284060955048, + "meanSquaredWeightError": 2.26705694217344e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005107489414513111, + "meanSquaredWeightError": 8.198658640878875e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001364879310131073, + "meanSquaredWeightError": 1.0674462203041912e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006516836583614349, + "meanSquaredWeightError": 9.449681925843834e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009464053437113762, + "meanSquaredWeightError": 4.404944853320103e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0265415720641613, + "meanSquaredWeightError": 8.238882287514571e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0333251990377903, + "meanSquaredWeightError": 2.927112973338808e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0194091796875, + "meanSquaredWeightError": 8.087501441877976e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005795937031507492, + "meanSquaredWeightError": 1.80652719450336e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005795937031507492, + "meanSquaredWeightError": 7.52482787191866e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0013129785656929016, + "meanSquaredWeightError": 9.38157196372913e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008208509534597397, + "meanSquaredWeightError": 8.128849593447285e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010572969913482666, + "meanSquaredWeightError": 4.412351817251192e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04150390625, + "meanSquaredWeightError": 8.620316975793685e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0302734375, + "meanSquaredWeightError": 2.6132290713576367e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02303466945886612, + "meanSquaredWeightError": 8.573984473514429e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006062665488570929, + "meanSquaredWeightError": 1.623409140449894e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005637351423501968, + "meanSquaredWeightError": 9.469766304448513e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008612200617790222, + "meanSquaredWeightError": 8.620569147410606e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008112378418445587, + "meanSquaredWeightError": 7.537758861531074e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008592978119850159, + "meanSquaredWeightError": 4.527719088542881e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.028881840407848358, + "meanSquaredWeightError": 9.06856655547017e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04156494140625, + "meanSquaredWeightError": 2.46402237280563e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01636962778866291, + "meanSquaredWeightError": 9.115400416703778e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02463379129767418, + "meanSquaredWeightError": 8.773882314017101e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.030029296875, + "meanSquaredWeightError": 1.9424107904342236e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02099609375, + "meanSquaredWeightError": 9.083250915864483e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00036885414738208055, + "meanSquaredWeightError": 3.3473028793196136e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005767084658145905, + "meanSquaredWeightError": 3.908651180495326e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001015007495880127, + "meanSquaredWeightError": 7.261003798220145e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005060629919171333, + "meanSquaredWeightError": 4.816435694721122e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005863220430910587, + "meanSquaredWeightError": 1.6181449069563314e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00045013427734375, + "meanSquaredWeightError": 6.057955026506079e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009198524057865143, + "meanSquaredWeightError": 8.711624310819843e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0013682451099157333, + "meanSquaredWeightError": 6.523535933666835e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008987076580524445, + "meanSquaredWeightError": 4.059202307615806e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.037841796875, + "meanSquaredWeightError": 8.265481596936297e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03264160454273224, + "meanSquaredWeightError": 1.6808718328320538e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0196533203125, + "meanSquaredWeightError": 8.899761496650171e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000939076766371727, + "meanSquaredWeightError": 2.0765920538678984e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005536414682865143, + "meanSquaredWeightError": 1.1359857943205043e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010246187448501587, + "meanSquaredWeightError": 7.989317651890815e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014816643670201302, + "meanSquaredWeightError": 6.387295137244564e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009112022817134857, + "meanSquaredWeightError": 3.856152730463691e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0225830078125, + "meanSquaredWeightError": 8.832107027956226e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02226562611758709, + "meanSquaredWeightError": 1.4517542012981721e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.017578125, + "meanSquaredWeightError": 9.189842558043892e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 254279680, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009068762883543968, + "meanSquaredWeightError": 1.1976288405435298e-08, + "storageBytes": 286064640 + } + ], + "quantizationTiers": [ + { + "scheme": "symmetric-int8", + "method": "rtn", + "groupSize": 32, + "include": [], + "exclude": [], + "matrices": 0 + }, + { + "scheme": "symmetric-int4", + "method": "gptq", + "groupSize": 32, + "include": [ + "p_model_model_language_model_layers_0_mlp_down_proj_weight", + "p_model_model_language_model_layers_0_mlp_gate_proj_weight", + "p_model_model_language_model_layers_0_mlp_up_proj_weight", + "p_model_model_language_model_layers_10_mlp_down_proj_weight", + "p_model_model_language_model_layers_10_mlp_gate_proj_weight", + "p_model_model_language_model_layers_10_mlp_up_proj_weight", + "p_model_model_language_model_layers_11_mlp_down_proj_weight", + "p_model_model_language_model_layers_11_mlp_gate_proj_weight", + "p_model_model_language_model_layers_11_mlp_up_proj_weight", + "p_model_model_language_model_layers_12_mlp_down_proj_weight", + "p_model_model_language_model_layers_12_mlp_gate_proj_weight", + "p_model_model_language_model_layers_12_mlp_up_proj_weight", + "p_model_model_language_model_layers_13_mlp_down_proj_weight", + "p_model_model_language_model_layers_13_mlp_gate_proj_weight", + "p_model_model_language_model_layers_13_mlp_up_proj_weight", + "p_model_model_language_model_layers_14_mlp_down_proj_weight", + "p_model_model_language_model_layers_14_mlp_gate_proj_weight", + "p_model_model_language_model_layers_14_mlp_up_proj_weight", + "p_model_model_language_model_layers_15_mlp_down_proj_weight", + "p_model_model_language_model_layers_15_mlp_gate_proj_weight", + "p_model_model_language_model_layers_15_mlp_up_proj_weight", + "p_model_model_language_model_layers_16_mlp_down_proj_weight", + "p_model_model_language_model_layers_16_mlp_gate_proj_weight", + "p_model_model_language_model_layers_16_mlp_up_proj_weight", + "p_model_model_language_model_layers_17_mlp_down_proj_weight", + "p_model_model_language_model_layers_17_mlp_gate_proj_weight", + "p_model_model_language_model_layers_17_mlp_up_proj_weight", + "p_model_model_language_model_layers_18_mlp_down_proj_weight", + "p_model_model_language_model_layers_18_mlp_gate_proj_weight", + "p_model_model_language_model_layers_18_mlp_up_proj_weight", + "p_model_model_language_model_layers_19_mlp_down_proj_weight", + "p_model_model_language_model_layers_19_mlp_gate_proj_weight", + "p_model_model_language_model_layers_19_mlp_up_proj_weight", + "p_model_model_language_model_layers_1_mlp_down_proj_weight", + "p_model_model_language_model_layers_1_mlp_gate_proj_weight", + "p_model_model_language_model_layers_1_mlp_up_proj_weight", + "p_model_model_language_model_layers_20_mlp_down_proj_weight", + "p_model_model_language_model_layers_20_mlp_gate_proj_weight", + "p_model_model_language_model_layers_20_mlp_up_proj_weight", + "p_model_model_language_model_layers_21_mlp_down_proj_weight", + "p_model_model_language_model_layers_21_mlp_gate_proj_weight", + "p_model_model_language_model_layers_21_mlp_up_proj_weight", + "p_model_model_language_model_layers_22_mlp_down_proj_weight", + "p_model_model_language_model_layers_22_mlp_gate_proj_weight", + "p_model_model_language_model_layers_22_mlp_up_proj_weight", + "p_model_model_language_model_layers_23_mlp_down_proj_weight", + "p_model_model_language_model_layers_23_mlp_gate_proj_weight", + "p_model_model_language_model_layers_23_mlp_up_proj_weight", + "p_model_model_language_model_layers_2_mlp_down_proj_weight", + "p_model_model_language_model_layers_2_mlp_gate_proj_weight", + "p_model_model_language_model_layers_2_mlp_up_proj_weight", + "p_model_model_language_model_layers_3_mlp_down_proj_weight", + "p_model_model_language_model_layers_3_mlp_gate_proj_weight", + "p_model_model_language_model_layers_3_mlp_up_proj_weight", + "p_model_model_language_model_layers_4_mlp_down_proj_weight", + "p_model_model_language_model_layers_4_mlp_gate_proj_weight", + "p_model_model_language_model_layers_4_mlp_up_proj_weight", + "p_model_model_language_model_layers_5_mlp_down_proj_weight", + "p_model_model_language_model_layers_5_mlp_gate_proj_weight", + "p_model_model_language_model_layers_5_mlp_up_proj_weight", + "p_model_model_language_model_layers_6_mlp_down_proj_weight", + "p_model_model_language_model_layers_6_mlp_gate_proj_weight", + "p_model_model_language_model_layers_6_mlp_up_proj_weight", + "p_model_model_language_model_layers_7_mlp_down_proj_weight", + "p_model_model_language_model_layers_7_mlp_gate_proj_weight", + "p_model_model_language_model_layers_7_mlp_up_proj_weight", + "p_model_model_language_model_layers_8_mlp_down_proj_weight", + "p_model_model_language_model_layers_8_mlp_gate_proj_weight", + "p_model_model_language_model_layers_8_mlp_up_proj_weight", + "p_model_model_language_model_layers_9_mlp_down_proj_weight", + "p_model_model_language_model_layers_9_mlp_gate_proj_weight", + "p_model_model_language_model_layers_9_mlp_up_proj_weight" + ], + "exclude": [], + "matrices": 72 + }, + { + "scheme": "symmetric-int8", + "method": "rtn", + "groupSize": 32, + "include": [ + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_11_self_attn_k_proj_weight", + "p_model_model_language_model_layers_11_self_attn_o_proj_weight", + "p_model_model_language_model_layers_11_self_attn_q_proj_weight", + "p_model_model_language_model_layers_11_self_attn_v_proj_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_15_self_attn_k_proj_weight", + "p_model_model_language_model_layers_15_self_attn_o_proj_weight", + "p_model_model_language_model_layers_15_self_attn_q_proj_weight", + "p_model_model_language_model_layers_15_self_attn_v_proj_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_19_self_attn_k_proj_weight", + "p_model_model_language_model_layers_19_self_attn_o_proj_weight", + "p_model_model_language_model_layers_19_self_attn_q_proj_weight", + "p_model_model_language_model_layers_19_self_attn_v_proj_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_23_self_attn_k_proj_weight", + "p_model_model_language_model_layers_23_self_attn_o_proj_weight", + "p_model_model_language_model_layers_23_self_attn_q_proj_weight", + "p_model_model_language_model_layers_23_self_attn_v_proj_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_3_self_attn_k_proj_weight", + "p_model_model_language_model_layers_3_self_attn_o_proj_weight", + "p_model_model_language_model_layers_3_self_attn_q_proj_weight", + "p_model_model_language_model_layers_3_self_attn_v_proj_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_7_self_attn_k_proj_weight", + "p_model_model_language_model_layers_7_self_attn_o_proj_weight", + "p_model_model_language_model_layers_7_self_attn_q_proj_weight", + "p_model_model_language_model_layers_7_self_attn_v_proj_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight" + ], + "exclude": [], + "matrices": 114 + } + ], + "states": [ + { + "input": "states_0", + "output": "slice_10", + "kind": "convolution" + }, + { + "input": "states_1", + "output": "slice_16", + "kind": "recurrent" + }, + { + "input": "states_2", + "output": "slice_18", + "kind": "convolution" + }, + { + "input": "states_3", + "output": "slice_24", + "kind": "recurrent" + }, + { + "input": "states_4", + "output": "slice_26", + "kind": "convolution" + }, + { + "input": "states_5", + "output": "slice_32", + "kind": "recurrent" + }, + { + "input": "states_6", + "output": "cache_update", + "kind": "kv" + }, + { + "input": "states_7", + "output": "cache_update_1", + "kind": "kv" + }, + { + "input": "states_8", + "output": "slice_44", + "kind": "convolution" + }, + { + "input": "states_9", + "output": "slice_50", + "kind": "recurrent" + }, + { + "input": "states_10", + "output": "slice_52", + "kind": "convolution" + }, + { + "input": "states_11", + "output": "slice_58", + "kind": "recurrent" + }, + { + "input": "states_12", + "output": "slice_60", + "kind": "convolution" + }, + { + "input": "states_13", + "output": "slice_66", + "kind": "recurrent" + }, + { + "input": "states_14", + "output": "cache_update_2", + "kind": "kv" + }, + { + "input": "states_15", + "output": "cache_update_3", + "kind": "kv" + }, + { + "input": "states_16", + "output": "slice_78", + "kind": "convolution" + }, + { + "input": "states_17", + "output": "slice_84", + "kind": "recurrent" + }, + { + "input": "states_18", + "output": "slice_86", + "kind": "convolution" + }, + { + "input": "states_19", + "output": "slice_92", + "kind": "recurrent" + }, + { + "input": "states_20", + "output": "slice_94", + "kind": "convolution" + }, + { + "input": "states_21", + "output": "slice_100", + "kind": "recurrent" + }, + { + "input": "states_22", + "output": "cache_update_4", + "kind": "kv" + }, + { + "input": "states_23", + "output": "cache_update_5", + "kind": "kv" + }, + { + "input": "states_24", + "output": "slice_112", + "kind": "convolution" + }, + { + "input": "states_25", + "output": "slice_118", + "kind": "recurrent" + }, + { + "input": "states_26", + "output": "slice_120", + "kind": "convolution" + }, + { + "input": "states_27", + "output": "slice_126", + "kind": "recurrent" + }, + { + "input": "states_28", + "output": "slice_128", + "kind": "convolution" + }, + { + "input": "states_29", + "output": "slice_134", + "kind": "recurrent" + }, + { + "input": "states_30", + "output": "cache_update_6", + "kind": "kv" + }, + { + "input": "states_31", + "output": "cache_update_7", + "kind": "kv" + }, + { + "input": "states_32", + "output": "slice_146", + "kind": "convolution" + }, + { + "input": "states_33", + "output": "slice_152", + "kind": "recurrent" + }, + { + "input": "states_34", + "output": "slice_154", + "kind": "convolution" + }, + { + "input": "states_35", + "output": "slice_160", + "kind": "recurrent" + }, + { + "input": "states_36", + "output": "slice_162", + "kind": "convolution" + }, + { + "input": "states_37", + "output": "slice_168", + "kind": "recurrent" + }, + { + "input": "states_38", + "output": "cache_update_8", + "kind": "kv" + }, + { + "input": "states_39", + "output": "cache_update_9", + "kind": "kv" + }, + { + "input": "states_40", + "output": "slice_180", + "kind": "convolution" + }, + { + "input": "states_41", + "output": "slice_186", + "kind": "recurrent" + }, + { + "input": "states_42", + "output": "slice_188", + "kind": "convolution" + }, + { + "input": "states_43", + "output": "slice_194", + "kind": "recurrent" + }, + { + "input": "states_44", + "output": "slice_196", + "kind": "convolution" + }, + { + "input": "states_45", + "output": "slice_202", + "kind": "recurrent" + }, + { + "input": "states_46", + "output": "cache_update_10", + "kind": "kv" + }, + { + "input": "states_47", + "output": "cache_update_11", + "kind": "kv" + } + ], + "dispatches": [ + { + "node": "embedding__part_0", + "inputs": [ + "input_ids", + "p_model_model_language_model_embed_tokens_weight__rows_0_131072", + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072" + ], + "output": "embedding__part_0", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/327049da924fdfe946e25d5d7e2cd46c6f5d3e801e1e42df106c86077a5354e9.wgsl" + }, + { + "node": "embedding__part_1", + "inputs": [ + "input_ids", + "p_model_model_language_model_embed_tokens_weight__rows_131072_117248", + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248" + ], + "output": "embedding__part_1", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/8239cfa752a60160251d979de4a5a3e2816d2f6af5094252d1a898e423ad5c46.wgsl" + }, + { + "node": "embedding", + "inputs": [ + "embedding__part_0", + "embedding__part_1" + ], + "output": "embedding", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "unsqueeze", + "inputs": [ + "position_ids" + ], + "output": "unsqueeze", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/18a66ec0875c456c9defd45d15e6ef0c148e64a1618e2825c7f96eadbbe8303a.wgsl" + }, + { + "node": "expand", + "inputs": [ + "unsqueeze" + ], + "output": "expand", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3caf74924884e14997a7fcbb2625a36e3b8a2bcf92da835bf6f787ddae437cdb.wgsl" + }, + { + "node": "slice_1", + "inputs": [ + "expand" + ], + "output": "slice_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/a9fa5db20fa8de1de429c6850c3e40ff8180bcdb59373919b84d2438b3b17352.wgsl" + }, + { + "node": "expand_1", + "inputs": [ + "b_model_model_language_model_rotary_emb_inv_freq" + ], + "output": "expand_1", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/a77f14c89662d2d3da8abd8f109667793dc662d307e5e1c61123be753e25437b.wgsl" + }, + { + "node": "_to_copy", + "inputs": [ + "slice_1" + ], + "output": "_to_copy", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/970aa8fd65c7e96611a4b6f30cb0b71e24bdaefff0ca117243cab1a5a7e645a9.wgsl" + }, + { + "node": "matmul", + "inputs": [ + "expand_1", + "_to_copy" + ], + "output": "matmul", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/1b9f4e330debc7235cfaf8f51d00fec575e09aa6116a45ea001e10ff5f09eb8c.wgsl" + }, + { + "node": "transpose", + "inputs": [ + "matmul" + ], + "output": "transpose", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/b12eae76f13225c0c5d35fc9309964aba3172712799ff3bd8538bb2c1b34aa33.wgsl" + }, + { + "node": "select", + "inputs": [ + "transpose" + ], + "output": "select", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/5f57b46016eccc3b50e3685e142b7deadce8f074f124e643e6e7015e2cc27357.wgsl" + }, + { + "node": "select_1", + "inputs": [ + "transpose" + ], + "output": "select_1", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/a1cf6b03baf18d389f0a2b95440c17125a8ecc1e633568de985d42cac6e0d1cb.wgsl" + }, + { + "node": "slice_2", + "inputs": [ + "select_1" + ], + "output": "slice_2", + "workgroups": [ + 3, + 1, + 1 + ], + "kernel": "kernels/3b8210d9acf57a3c1cca9e926547520d5994967f27ce92897a15b14137f7da66.wgsl" + }, + { + "node": "slice_3", + "inputs": [ + "select" + ], + "output": "slice_3", + "workgroups": [ + 3, + 1, + 1 + ], + "kernel": "kernels/3b8210d9acf57a3c1cca9e926547520d5994967f27ce92897a15b14137f7da66.wgsl" + }, + { + "node": "copy", + "inputs": [ + "slice_2" + ], + "output": "copy", + "workgroups": [ + 3, + 1, + 1 + ], + "kernel": "kernels/d44354e74d5bfca2fb879607683c5d04cb30c4e1a694270584a01c0c6b3d9be8.wgsl" + }, + { + "node": "transpose_1", + "inputs": [ + "matmul" + ], + "output": "transpose_1", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/b12eae76f13225c0c5d35fc9309964aba3172712799ff3bd8538bb2c1b34aa33.wgsl" + }, + { + "node": "select_2", + "inputs": [ + "transpose_1" + ], + "output": "select_2", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/5f57b46016eccc3b50e3685e142b7deadce8f074f124e643e6e7015e2cc27357.wgsl" + }, + { + "node": "slice_scatter", + "inputs": [ + "copy", + "select_2" + ], + "output": "slice_scatter", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/3daeb9cba5c07ed47ac55e1c2565aa82f7d8e8409037707e7339d1d96d122144.wgsl" + }, + { + "node": "select_scatter", + "inputs": [ + "slice_scatter", + "transpose_1" + ], + "output": "select_scatter", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/4fa24e8f09b07271534341707a7a688af67c7deca1f114bfaeba20a2e066c9f1.wgsl" + }, + { + "node": "transpose_2", + "inputs": [ + "select_scatter" + ], + "output": "transpose_2", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/d155dd8abb0fe375aea320241d54486b2489a35ad76150cc77c4778bfbb9d957.wgsl" + }, + { + "node": "transpose_7", + "inputs": [ + "transpose_2" + ], + "output": "transpose_7", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/b12eae76f13225c0c5d35fc9309964aba3172712799ff3bd8538bb2c1b34aa33.wgsl" + }, + { + "node": "select_7", + "inputs": [ + "transpose_7" + ], + "output": "select_7", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/5f57b46016eccc3b50e3685e142b7deadce8f074f124e643e6e7015e2cc27357.wgsl" + }, + { + "node": "slice_7", + "inputs": [ + "select_7" + ], + "output": "slice_7", + "workgroups": [ + 3, + 1, + 1 + ], + "kernel": "kernels/5fbfe4534ce5bfff645563250dd536998f537f4d690c4c2fc422723d5af8e3ec.wgsl" + }, + { + "node": "transpose_8", + "inputs": [ + "transpose_2" + ], + "output": "transpose_8", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/b12eae76f13225c0c5d35fc9309964aba3172712799ff3bd8538bb2c1b34aa33.wgsl" + }, + { + "node": "select_8", + "inputs": [ + "transpose_8" + ], + "output": "select_8", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/deb451a300a279d0325fe3be090515d4daae6b3232d87b8e2c050991388ac00f.wgsl" + }, + { + "node": "slice_8", + "inputs": [ + "select_8" + ], + "output": "slice_8", + "workgroups": [ + 3, + 1, + 1 + ], + "kernel": "kernels/5fbfe4534ce5bfff645563250dd536998f537f4d690c4c2fc422723d5af8e3ec.wgsl" + }, + { + "node": "copy_1", + "inputs": [ + "slice_8" + ], + "output": "copy_1", + "workgroups": [ + 3, + 1, + 1 + ], + "kernel": "kernels/c2e7cdc03edb25ef7a160965c17c559b6ec7e139c145dc7390b73523bacba5bb.wgsl" + }, + { + "node": "transpose_9", + "inputs": [ + "transpose_2" + ], + "output": "transpose_9", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/b12eae76f13225c0c5d35fc9309964aba3172712799ff3bd8538bb2c1b34aa33.wgsl" + }, + { + "node": "select_9", + "inputs": [ + "transpose_9" + ], + "output": "select_9", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/5f57b46016eccc3b50e3685e142b7deadce8f074f124e643e6e7015e2cc27357.wgsl" + }, + { + "node": "slice_scatter_1", + "inputs": [ + "copy_1", + "select_9" + ], + "output": "slice_scatter_1", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/359d61f2ba174e54e2a0920fcf61e84e436fd3eaf7e2832e21fdb24094493667.wgsl" + }, + { + "node": "select_scatter_1", + "inputs": [ + "slice_scatter_1", + "transpose_9" + ], + "output": "select_scatter_1", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/4fa24e8f09b07271534341707a7a688af67c7deca1f114bfaeba20a2e066c9f1.wgsl" + }, + { + "node": "transpose_10", + "inputs": [ + "select_scatter_1" + ], + "output": "transpose_10", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/d155dd8abb0fe375aea320241d54486b2489a35ad76150cc77c4778bfbb9d957.wgsl" + }, + { + "node": "transpose_12", + "inputs": [ + "transpose_10" + ], + "output": "transpose_12", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/b12eae76f13225c0c5d35fc9309964aba3172712799ff3bd8538bb2c1b34aa33.wgsl" + }, + { + "node": "select_11", + "inputs": [ + "transpose_12" + ], + "output": "select_11", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/5f57b46016eccc3b50e3685e142b7deadce8f074f124e643e6e7015e2cc27357.wgsl" + }, + { + "node": "cat", + "inputs": [ + "select_11" + ], + "output": "cat", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/1a73b6fceda28f40bba4ae27a986ffa39842db94fa5903ea5bf689a7e854d4f3.wgsl" + }, + { + "node": "cos", + "inputs": [ + "cat" + ], + "output": "cos", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/764403c33cd8994568e068810380cb57e9f2e326dfc09bed63eed1fbed044da1.wgsl" + }, + { + "node": "mul", + "inputs": [ + "cos" + ], + "output": "mul", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9b7697d3e1de37277befbd725307f40a5143faae3dcf43182d4ad89d8572ed42.wgsl" + }, + { + "node": "sin", + "inputs": [ + "cat" + ], + "output": "sin", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/ca9a3115f8f3881eee99e421cb4f259113b13bb4c107348279bb92c643a52197.wgsl" + }, + { + "node": "mul_1", + "inputs": [ + "sin" + ], + "output": "mul_1", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9b7697d3e1de37277befbd725307f40a5143faae3dcf43182d4ad89d8572ed42.wgsl" + }, + { + "node": "mul_3", + "inputs": [ + "embedding", + "p_model_model_language_model_layers_0_input_layernorm_weight" + ], + "output": "mul_3", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear", + "inputs": [ + "mul_3", + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_13", + "inputs": [ + "linear" + ], + "output": "transpose_13", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_1", + "inputs": [ + "mul_3", + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_1", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_2", + "inputs": [ + "mul_3", + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_3", + "inputs": [ + "mul_3", + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_1", + "inputs": [ + "states_0", + "transpose_13" + ], + "output": "cat_1", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_10", + "inputs": [ + "cat_1" + ], + "output": "slice_10", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_0" + }, + { + "node": "conv1d", + "inputs": [ + "cat_1", + "p_model_model_language_model_layers_0_linear_attn_conv1d_weight" + ], + "output": "conv1d", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_11", + "inputs": [ + "conv1d" + ], + "output": "slice_11", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu", + "inputs": [ + "slice_11" + ], + "output": "silu", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_12", + "inputs": [ + "silu" + ], + "output": "slice_12", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_14", + "inputs": [ + "slice_12" + ], + "output": "transpose_14", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_13", + "inputs": [ + "transpose_14" + ], + "output": "slice_13", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_14", + "inputs": [ + "transpose_14" + ], + "output": "slice_14", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_15", + "inputs": [ + "transpose_14" + ], + "output": "slice_15", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid", + "inputs": [ + "linear_2" + ], + "output": "sigmoid", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp", + "inputs": [ + "p_model_model_language_model_layers_0_linear_attn_a_log" + ], + "output": "exp", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg", + "inputs": [ + "exp" + ], + "output": "neg", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_2", + "inputs": [ + "linear_3", + "p_model_model_language_model_layers_0_linear_attn_dt_bias" + ], + "output": "add_2", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus", + "inputs": [ + "add_2" + ], + "output": "softplus", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_4", + "inputs": [ + "neg", + "softplus" + ], + "output": "mul_4", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_5", + "inputs": [ + "slice_13" + ], + "output": "mul_5", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_1", + "inputs": [ + "mul_5" + ], + "output": "sum_1", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_3", + "inputs": [ + "sum_1" + ], + "output": "add_3", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_1", + "inputs": [ + "add_3" + ], + "output": "rsqrt_1", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_6", + "inputs": [ + "slice_13", + "rsqrt_1" + ], + "output": "mul_6", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_7", + "inputs": [ + "slice_14" + ], + "output": "mul_7", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_2", + "inputs": [ + "mul_7" + ], + "output": "sum_2", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_4", + "inputs": [ + "sum_2" + ], + "output": "add_4", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_2", + "inputs": [ + "add_4" + ], + "output": "rsqrt_2", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_8", + "inputs": [ + "slice_14", + "rsqrt_2" + ], + "output": "mul_8", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule", + "inputs": [ + "mul_6", + "mul_8", + "slice_15", + "mul_4", + "sigmoid", + "states_1" + ], + "output": "normalized_gated_delta_rule", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_16", + "inputs": [ + "normalized_gated_delta_rule" + ], + "output": "slice_16", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_1" + }, + { + "node": "slice_17", + "inputs": [ + "normalized_gated_delta_rule" + ], + "output": "slice_17", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_15", + "inputs": [ + "slice_17" + ], + "output": "transpose_15", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_10", + "inputs": [ + "transpose_15", + "p_model_model_language_model_layers_0_linear_attn_norm_weight" + ], + "output": "mul_10", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_11", + "inputs": [ + "linear_1", + "mul_10" + ], + "output": "mul_11", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_4", + "inputs": [ + "mul_11", + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_4", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_6", + "inputs": [ + "embedding", + "linear_4" + ], + "output": "add_6", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_13", + "inputs": [ + "add_6", + "p_model_model_language_model_layers_0_post_attention_layernorm_weight" + ], + "output": "mul_13", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_5", + "inputs": [ + "mul_13", + "p_model_model_language_model_layers_0_mlp_gate_proj_weight", + "p_model_model_language_model_layers_0_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_5", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_6", + "inputs": [ + "mul_13", + "p_model_model_language_model_layers_0_mlp_up_proj_weight", + "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_6", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_14", + "inputs": [ + "linear_5", + "linear_6" + ], + "output": "mul_14", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_7", + "inputs": [ + "mul_14", + "p_model_model_language_model_layers_0_mlp_down_proj_weight", + "p_model_model_language_model_layers_0_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_7", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_9", + "inputs": [ + "add_6", + "linear_7" + ], + "output": "add_9", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_16", + "inputs": [ + "add_9", + "p_model_model_language_model_layers_1_input_layernorm_weight" + ], + "output": "mul_16", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_8", + "inputs": [ + "mul_16", + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_8", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_16", + "inputs": [ + "linear_8" + ], + "output": "transpose_16", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_9", + "inputs": [ + "mul_16", + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_9", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_10", + "inputs": [ + "mul_16", + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_11", + "inputs": [ + "mul_16", + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_2", + "inputs": [ + "states_2", + "transpose_16" + ], + "output": "cat_2", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_18", + "inputs": [ + "cat_2" + ], + "output": "slice_18", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_2" + }, + { + "node": "conv1d_1", + "inputs": [ + "cat_2", + "p_model_model_language_model_layers_1_linear_attn_conv1d_weight" + ], + "output": "conv1d_1", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_19", + "inputs": [ + "conv1d_1" + ], + "output": "slice_19", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_3", + "inputs": [ + "slice_19" + ], + "output": "silu_3", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_20", + "inputs": [ + "silu_3" + ], + "output": "slice_20", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_17", + "inputs": [ + "slice_20" + ], + "output": "transpose_17", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_21", + "inputs": [ + "transpose_17" + ], + "output": "slice_21", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_22", + "inputs": [ + "transpose_17" + ], + "output": "slice_22", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_23", + "inputs": [ + "transpose_17" + ], + "output": "slice_23", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_1", + "inputs": [ + "linear_10" + ], + "output": "sigmoid_1", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_1", + "inputs": [ + "p_model_model_language_model_layers_1_linear_attn_a_log" + ], + "output": "exp_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_1", + "inputs": [ + "exp_1" + ], + "output": "neg_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_12", + "inputs": [ + "linear_11", + "p_model_model_language_model_layers_1_linear_attn_dt_bias" + ], + "output": "add_12", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_1", + "inputs": [ + "add_12" + ], + "output": "softplus_1", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_17", + "inputs": [ + "neg_1", + "softplus_1" + ], + "output": "mul_17", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_18", + "inputs": [ + "slice_21" + ], + "output": "mul_18", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_3", + "inputs": [ + "mul_18" + ], + "output": "sum_3", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_13", + "inputs": [ + "sum_3" + ], + "output": "add_13", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_6", + "inputs": [ + "add_13" + ], + "output": "rsqrt_6", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_19", + "inputs": [ + "slice_21", + "rsqrt_6" + ], + "output": "mul_19", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_20", + "inputs": [ + "slice_22" + ], + "output": "mul_20", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_4", + "inputs": [ + "mul_20" + ], + "output": "sum_4", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_14", + "inputs": [ + "sum_4" + ], + "output": "add_14", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_7", + "inputs": [ + "add_14" + ], + "output": "rsqrt_7", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_21", + "inputs": [ + "slice_22", + "rsqrt_7" + ], + "output": "mul_21", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_1", + "inputs": [ + "mul_19", + "mul_21", + "slice_23", + "mul_17", + "sigmoid_1", + "states_3" + ], + "output": "normalized_gated_delta_rule_1", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_24", + "inputs": [ + "normalized_gated_delta_rule_1" + ], + "output": "slice_24", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_3" + }, + { + "node": "slice_25", + "inputs": [ + "normalized_gated_delta_rule_1" + ], + "output": "slice_25", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_18", + "inputs": [ + "slice_25" + ], + "output": "transpose_18", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_23", + "inputs": [ + "transpose_18", + "p_model_model_language_model_layers_1_linear_attn_norm_weight" + ], + "output": "mul_23", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_24", + "inputs": [ + "linear_9", + "mul_23" + ], + "output": "mul_24", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_12", + "inputs": [ + "mul_24", + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_12", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_16", + "inputs": [ + "add_9", + "linear_12" + ], + "output": "add_16", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_26", + "inputs": [ + "add_16", + "p_model_model_language_model_layers_1_post_attention_layernorm_weight" + ], + "output": "mul_26", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_13", + "inputs": [ + "mul_26", + "p_model_model_language_model_layers_1_mlp_gate_proj_weight", + "p_model_model_language_model_layers_1_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_13", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_14", + "inputs": [ + "mul_26", + "p_model_model_language_model_layers_1_mlp_up_proj_weight", + "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_14", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_27", + "inputs": [ + "linear_13", + "linear_14" + ], + "output": "mul_27", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_15", + "inputs": [ + "mul_27", + "p_model_model_language_model_layers_1_mlp_down_proj_weight", + "p_model_model_language_model_layers_1_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_15", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_19", + "inputs": [ + "add_16", + "linear_15" + ], + "output": "add_19", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_29", + "inputs": [ + "add_19", + "p_model_model_language_model_layers_2_input_layernorm_weight" + ], + "output": "mul_29", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_16", + "inputs": [ + "mul_29", + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_16", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_19", + "inputs": [ + "linear_16" + ], + "output": "transpose_19", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_17", + "inputs": [ + "mul_29", + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_17", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_18", + "inputs": [ + "mul_29", + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_18", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_19", + "inputs": [ + "mul_29", + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_19", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_3", + "inputs": [ + "states_4", + "transpose_19" + ], + "output": "cat_3", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_26", + "inputs": [ + "cat_3" + ], + "output": "slice_26", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_4" + }, + { + "node": "conv1d_2", + "inputs": [ + "cat_3", + "p_model_model_language_model_layers_2_linear_attn_conv1d_weight" + ], + "output": "conv1d_2", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_27", + "inputs": [ + "conv1d_2" + ], + "output": "slice_27", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_6", + "inputs": [ + "slice_27" + ], + "output": "silu_6", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_28", + "inputs": [ + "silu_6" + ], + "output": "slice_28", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_20", + "inputs": [ + "slice_28" + ], + "output": "transpose_20", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_29", + "inputs": [ + "transpose_20" + ], + "output": "slice_29", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_30", + "inputs": [ + "transpose_20" + ], + "output": "slice_30", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_31", + "inputs": [ + "transpose_20" + ], + "output": "slice_31", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_2", + "inputs": [ + "linear_18" + ], + "output": "sigmoid_2", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_2", + "inputs": [ + "p_model_model_language_model_layers_2_linear_attn_a_log" + ], + "output": "exp_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_2", + "inputs": [ + "exp_2" + ], + "output": "neg_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_22", + "inputs": [ + "linear_19", + "p_model_model_language_model_layers_2_linear_attn_dt_bias" + ], + "output": "add_22", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_2", + "inputs": [ + "add_22" + ], + "output": "softplus_2", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_30", + "inputs": [ + "neg_2", + "softplus_2" + ], + "output": "mul_30", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_31", + "inputs": [ + "slice_29" + ], + "output": "mul_31", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_5", + "inputs": [ + "mul_31" + ], + "output": "sum_5", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_23", + "inputs": [ + "sum_5" + ], + "output": "add_23", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_11", + "inputs": [ + "add_23" + ], + "output": "rsqrt_11", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_32", + "inputs": [ + "slice_29", + "rsqrt_11" + ], + "output": "mul_32", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_33", + "inputs": [ + "slice_30" + ], + "output": "mul_33", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_6", + "inputs": [ + "mul_33" + ], + "output": "sum_6", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_24", + "inputs": [ + "sum_6" + ], + "output": "add_24", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_12", + "inputs": [ + "add_24" + ], + "output": "rsqrt_12", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_34", + "inputs": [ + "slice_30", + "rsqrt_12" + ], + "output": "mul_34", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_2", + "inputs": [ + "mul_32", + "mul_34", + "slice_31", + "mul_30", + "sigmoid_2", + "states_5" + ], + "output": "normalized_gated_delta_rule_2", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_32", + "inputs": [ + "normalized_gated_delta_rule_2" + ], + "output": "slice_32", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_5" + }, + { + "node": "slice_33", + "inputs": [ + "normalized_gated_delta_rule_2" + ], + "output": "slice_33", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_21", + "inputs": [ + "slice_33" + ], + "output": "transpose_21", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_36", + "inputs": [ + "transpose_21", + "p_model_model_language_model_layers_2_linear_attn_norm_weight" + ], + "output": "mul_36", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_37", + "inputs": [ + "linear_17", + "mul_36" + ], + "output": "mul_37", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_20", + "inputs": [ + "mul_37", + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_20", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_26", + "inputs": [ + "add_19", + "linear_20" + ], + "output": "add_26", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_39", + "inputs": [ + "add_26", + "p_model_model_language_model_layers_2_post_attention_layernorm_weight" + ], + "output": "mul_39", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_21", + "inputs": [ + "mul_39", + "p_model_model_language_model_layers_2_mlp_gate_proj_weight", + "p_model_model_language_model_layers_2_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_21", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_22", + "inputs": [ + "mul_39", + "p_model_model_language_model_layers_2_mlp_up_proj_weight", + "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_22", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_40", + "inputs": [ + "linear_21", + "linear_22" + ], + "output": "mul_40", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_23", + "inputs": [ + "mul_40", + "p_model_model_language_model_layers_2_mlp_down_proj_weight", + "p_model_model_language_model_layers_2_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_23", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_29", + "inputs": [ + "add_26", + "linear_23" + ], + "output": "add_29", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_42", + "inputs": [ + "add_29", + "p_model_model_language_model_layers_3_input_layernorm_weight" + ], + "output": "mul_42", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_24", + "inputs": [ + "mul_42", + "p_model_model_language_model_layers_3_self_attn_q_proj_weight", + "p_model_model_language_model_layers_3_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_24", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/8b540ef0dbe320e8ccc513d0b118e62c6d25dff607a62b8d9de03d1a393a0a5c.wgsl" + }, + { + "node": "slice_34", + "inputs": [ + "linear_24" + ], + "output": "slice_34", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/9d3d7b777701cd320ca11358e554e1019fb041ee4d765e66071c215242906773.wgsl" + }, + { + "node": "slice_35", + "inputs": [ + "linear_24" + ], + "output": "slice_35", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/3622ffa2072335a35d6c9db5d284ed92b3fb5ed304acc2c606408063e7e21dcd.wgsl" + }, + { + "node": "mul_44", + "inputs": [ + "slice_34", + "p_model_model_language_model_layers_3_self_attn_q_norm_weight" + ], + "output": "mul_44", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl" + }, + { + "node": "transpose_22", + "inputs": [ + "mul_44" + ], + "output": "transpose_22", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/63396da046b345383e1c7b09aee551d08e4c3c45030ff5c8bb76bbc34bb2abc6.wgsl" + }, + { + "node": "linear_25", + "inputs": [ + "mul_42", + "p_model_model_language_model_layers_3_self_attn_k_proj_weight", + "p_model_model_language_model_layers_3_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_25", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl" + }, + { + "node": "mul_46", + "inputs": [ + "linear_25", + "p_model_model_language_model_layers_3_self_attn_k_norm_weight" + ], + "output": "mul_46", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl" + }, + { + "node": "transpose_23", + "inputs": [ + "mul_46" + ], + "output": "transpose_23", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl" + }, + { + "node": "linear_26", + "inputs": [ + "mul_42", + "p_model_model_language_model_layers_3_self_attn_v_proj_weight", + "p_model_model_language_model_layers_3_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_26", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl" + }, + { + "node": "transpose_24", + "inputs": [ + "linear_26" + ], + "output": "transpose_24", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl" + }, + { + "node": "slice_36", + "inputs": [ + "transpose_22" + ], + "output": "slice_36", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f2ee90b3d949141c23610237d27d44a7b833aacae8cdbc2e42a595d5eb7b5537.wgsl" + }, + { + "node": "slice_37", + "inputs": [ + "transpose_22" + ], + "output": "slice_37", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/8d62df48089497dd19fea48680730210decda42b63f69ecdf9688fc4a43efc58.wgsl" + }, + { + "node": "slice_38", + "inputs": [ + "transpose_23" + ], + "output": "slice_38", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6caa35fcab56c639d1859e1b090a7e5bf016dd7f33ecd2373c24ab625de16e09.wgsl" + }, + { + "node": "slice_39", + "inputs": [ + "transpose_23" + ], + "output": "slice_39", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/2182574f59b7d9e38d74b30540c5781827e7da2de1ba6b00327af0b4cb42379d.wgsl" + }, + { + "node": "mul_47", + "inputs": [ + "slice_36", + "mul" + ], + "output": "mul_47", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e3c6a64e01b6ea79adf125ff00a6dcf00f0b65aee0f4a3d14fdbc20e5dfeb22d.wgsl" + }, + { + "node": "slice_40", + "inputs": [ + "slice_36" + ], + "output": "slice_40", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/b5984ff6eb9e56ef0f9930f775c1dc3c203eec15924e5fea4c2525b2ce180857.wgsl" + }, + { + "node": "slice_41", + "inputs": [ + "slice_36" + ], + "output": "slice_41", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/8256ea6cdc3500306470c0bb2dcdb44ebdd454dafa218ad6a94dd953b9bbe125.wgsl" + }, + { + "node": "neg_3", + "inputs": [ + "slice_41" + ], + "output": "neg_3", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/adf1e58a124e1301625024ae142dca5fb1dd50799fa94b52eaeffb82ea99852c.wgsl" + }, + { + "node": "cat_4", + "inputs": [ + "neg_3", + "slice_40" + ], + "output": "cat_4", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl" + }, + { + "node": "mul_48", + "inputs": [ + "cat_4", + "mul_1" + ], + "output": "mul_48", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e3c6a64e01b6ea79adf125ff00a6dcf00f0b65aee0f4a3d14fdbc20e5dfeb22d.wgsl" + }, + { + "node": "add_36", + "inputs": [ + "mul_47", + "mul_48" + ], + "output": "add_36", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl" + }, + { + "node": "mul_49", + "inputs": [ + "slice_38", + "mul" + ], + "output": "mul_49", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl" + }, + { + "node": "slice_42", + "inputs": [ + "slice_38" + ], + "output": "slice_42", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/58077eb80a5fa253f535c97c82dc07edd48455a17cc52583aeee41c7d1bbd0b3.wgsl" + }, + { + "node": "slice_43", + "inputs": [ + "slice_38" + ], + "output": "slice_43", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3830061b0fc54c85d3d8a783b927c44484b1a3b68b52b8d8a1b980f7553deebf.wgsl" + }, + { + "node": "neg_4", + "inputs": [ + "slice_43" + ], + "output": "neg_4", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl" + }, + { + "node": "cat_5", + "inputs": [ + "neg_4", + "slice_42" + ], + "output": "cat_5", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/bda7ebd3bcf25c31938fca1a582e82b972ab742f85f826f085417b0e6f651e60.wgsl" + }, + { + "node": "mul_50", + "inputs": [ + "cat_5", + "mul_1" + ], + "output": "mul_50", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl" + }, + { + "node": "add_37", + "inputs": [ + "mul_49", + "mul_50" + ], + "output": "add_37", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl" + }, + { + "node": "cat_6", + "inputs": [ + "add_36", + "slice_37" + ], + "output": "cat_6", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl" + }, + { + "node": "cat_7", + "inputs": [ + "add_37", + "slice_39" + ], + "output": "cat_7", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl" + }, + { + "node": "cache_update", + "inputs": [ + "cat_7", + "position_ids" + ], + "output": "cache_update", + "workgroups": [ + 128, + 1, + 1 + ], + "inPlaceOf": "states_6", + "kernel": "kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl" + }, + { + "node": "cache_update_1", + "inputs": [ + "transpose_24", + "position_ids" + ], + "output": "cache_update_1", + "workgroups": [ + 128, + 1, + 1 + ], + "inPlaceOf": "states_7", + "kernel": "kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl" + }, + { + "node": "matmul_2", + "inputs": [ + "cat_6", + "cache_update", + "cache_update_1", + "attention_mask" + ], + "output": "matmul_2", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d9ab25f0ef4248028d7dab1e782d0756dfc664cc90794d13644bd91106541608.wgsl" + }, + { + "node": "transpose_26", + "inputs": [ + "matmul_2" + ], + "output": "transpose_26", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/84f24861effe9e2f3b05f98fe6d4e16f4f8204be40a7eeece0668e57060ecc7b.wgsl" + }, + { + "node": "sigmoid_3", + "inputs": [ + "slice_35" + ], + "output": "sigmoid_3", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/816861e23c92e156c1ac01f21e4f6dd1cbfc14978b1be8a5994fb45b0f845cb8.wgsl" + }, + { + "node": "mul_52", + "inputs": [ + "transpose_26", + "sigmoid_3" + ], + "output": "mul_52", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/aa227ae49fab3f75d86e02e6f390be785eabfeafe70e77a5c7eb4ad51895e382.wgsl" + }, + { + "node": "linear_27", + "inputs": [ + "mul_52", + "p_model_model_language_model_layers_3_self_attn_o_proj_weight", + "p_model_model_language_model_layers_3_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_27", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_39", + "inputs": [ + "add_29", + "linear_27" + ], + "output": "add_39", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_54", + "inputs": [ + "add_39", + "p_model_model_language_model_layers_3_post_attention_layernorm_weight" + ], + "output": "mul_54", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_28", + "inputs": [ + "mul_54", + "p_model_model_language_model_layers_3_mlp_gate_proj_weight", + "p_model_model_language_model_layers_3_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_28", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_29", + "inputs": [ + "mul_54", + "p_model_model_language_model_layers_3_mlp_up_proj_weight", + "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_29", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_55", + "inputs": [ + "linear_28", + "linear_29" + ], + "output": "mul_55", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_30", + "inputs": [ + "mul_55", + "p_model_model_language_model_layers_3_mlp_down_proj_weight", + "p_model_model_language_model_layers_3_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_30", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_42", + "inputs": [ + "add_39", + "linear_30" + ], + "output": "add_42", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_57", + "inputs": [ + "add_42", + "p_model_model_language_model_layers_4_input_layernorm_weight" + ], + "output": "mul_57", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_31", + "inputs": [ + "mul_57", + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_31", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_27", + "inputs": [ + "linear_31" + ], + "output": "transpose_27", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_32", + "inputs": [ + "mul_57", + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_32", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_33", + "inputs": [ + "mul_57", + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_33", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_34", + "inputs": [ + "mul_57", + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_34", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_8", + "inputs": [ + "states_8", + "transpose_27" + ], + "output": "cat_8", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_44", + "inputs": [ + "cat_8" + ], + "output": "slice_44", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_8" + }, + { + "node": "conv1d_3", + "inputs": [ + "cat_8", + "p_model_model_language_model_layers_4_linear_attn_conv1d_weight" + ], + "output": "conv1d_3", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_45", + "inputs": [ + "conv1d_3" + ], + "output": "slice_45", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_10", + "inputs": [ + "slice_45" + ], + "output": "silu_10", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_46", + "inputs": [ + "silu_10" + ], + "output": "slice_46", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_28", + "inputs": [ + "slice_46" + ], + "output": "transpose_28", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_47", + "inputs": [ + "transpose_28" + ], + "output": "slice_47", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_48", + "inputs": [ + "transpose_28" + ], + "output": "slice_48", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_49", + "inputs": [ + "transpose_28" + ], + "output": "slice_49", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_4", + "inputs": [ + "linear_33" + ], + "output": "sigmoid_4", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_3", + "inputs": [ + "p_model_model_language_model_layers_4_linear_attn_a_log" + ], + "output": "exp_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_5", + "inputs": [ + "exp_3" + ], + "output": "neg_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_45", + "inputs": [ + "linear_34", + "p_model_model_language_model_layers_4_linear_attn_dt_bias" + ], + "output": "add_45", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_3", + "inputs": [ + "add_45" + ], + "output": "softplus_3", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_58", + "inputs": [ + "neg_5", + "softplus_3" + ], + "output": "mul_58", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_59", + "inputs": [ + "slice_47" + ], + "output": "mul_59", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_7", + "inputs": [ + "mul_59" + ], + "output": "sum_7", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_46", + "inputs": [ + "sum_7" + ], + "output": "add_46", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_20", + "inputs": [ + "add_46" + ], + "output": "rsqrt_20", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_60", + "inputs": [ + "slice_47", + "rsqrt_20" + ], + "output": "mul_60", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_61", + "inputs": [ + "slice_48" + ], + "output": "mul_61", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_8", + "inputs": [ + "mul_61" + ], + "output": "sum_8", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_47", + "inputs": [ + "sum_8" + ], + "output": "add_47", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_21", + "inputs": [ + "add_47" + ], + "output": "rsqrt_21", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_62", + "inputs": [ + "slice_48", + "rsqrt_21" + ], + "output": "mul_62", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_3", + "inputs": [ + "mul_60", + "mul_62", + "slice_49", + "mul_58", + "sigmoid_4", + "states_9" + ], + "output": "normalized_gated_delta_rule_3", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_50", + "inputs": [ + "normalized_gated_delta_rule_3" + ], + "output": "slice_50", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_9" + }, + { + "node": "slice_51", + "inputs": [ + "normalized_gated_delta_rule_3" + ], + "output": "slice_51", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_29", + "inputs": [ + "slice_51" + ], + "output": "transpose_29", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_64", + "inputs": [ + "transpose_29", + "p_model_model_language_model_layers_4_linear_attn_norm_weight" + ], + "output": "mul_64", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_65", + "inputs": [ + "linear_32", + "mul_64" + ], + "output": "mul_65", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_35", + "inputs": [ + "mul_65", + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_35", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_49", + "inputs": [ + "add_42", + "linear_35" + ], + "output": "add_49", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_67", + "inputs": [ + "add_49", + "p_model_model_language_model_layers_4_post_attention_layernorm_weight" + ], + "output": "mul_67", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_36", + "inputs": [ + "mul_67", + "p_model_model_language_model_layers_4_mlp_gate_proj_weight", + "p_model_model_language_model_layers_4_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_36", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_37", + "inputs": [ + "mul_67", + "p_model_model_language_model_layers_4_mlp_up_proj_weight", + "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_37", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_68", + "inputs": [ + "linear_36", + "linear_37" + ], + "output": "mul_68", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_38", + "inputs": [ + "mul_68", + "p_model_model_language_model_layers_4_mlp_down_proj_weight", + "p_model_model_language_model_layers_4_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_38", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_52", + "inputs": [ + "add_49", + "linear_38" + ], + "output": "add_52", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_70", + "inputs": [ + "add_52", + "p_model_model_language_model_layers_5_input_layernorm_weight" + ], + "output": "mul_70", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_39", + "inputs": [ + "mul_70", + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_39", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_30", + "inputs": [ + "linear_39" + ], + "output": "transpose_30", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_40", + "inputs": [ + "mul_70", + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_40", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_41", + "inputs": [ + "mul_70", + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_41", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_42", + "inputs": [ + "mul_70", + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_42", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_9", + "inputs": [ + "states_10", + "transpose_30" + ], + "output": "cat_9", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_52", + "inputs": [ + "cat_9" + ], + "output": "slice_52", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_10" + }, + { + "node": "conv1d_4", + "inputs": [ + "cat_9", + "p_model_model_language_model_layers_5_linear_attn_conv1d_weight" + ], + "output": "conv1d_4", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_53", + "inputs": [ + "conv1d_4" + ], + "output": "slice_53", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_13", + "inputs": [ + "slice_53" + ], + "output": "silu_13", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_54", + "inputs": [ + "silu_13" + ], + "output": "slice_54", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_31", + "inputs": [ + "slice_54" + ], + "output": "transpose_31", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_55", + "inputs": [ + "transpose_31" + ], + "output": "slice_55", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_56", + "inputs": [ + "transpose_31" + ], + "output": "slice_56", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_57", + "inputs": [ + "transpose_31" + ], + "output": "slice_57", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_5", + "inputs": [ + "linear_41" + ], + "output": "sigmoid_5", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_4", + "inputs": [ + "p_model_model_language_model_layers_5_linear_attn_a_log" + ], + "output": "exp_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_6", + "inputs": [ + "exp_4" + ], + "output": "neg_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_55", + "inputs": [ + "linear_42", + "p_model_model_language_model_layers_5_linear_attn_dt_bias" + ], + "output": "add_55", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_4", + "inputs": [ + "add_55" + ], + "output": "softplus_4", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_71", + "inputs": [ + "neg_6", + "softplus_4" + ], + "output": "mul_71", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_72", + "inputs": [ + "slice_55" + ], + "output": "mul_72", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_9", + "inputs": [ + "mul_72" + ], + "output": "sum_9", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_56", + "inputs": [ + "sum_9" + ], + "output": "add_56", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_25", + "inputs": [ + "add_56" + ], + "output": "rsqrt_25", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_73", + "inputs": [ + "slice_55", + "rsqrt_25" + ], + "output": "mul_73", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_74", + "inputs": [ + "slice_56" + ], + "output": "mul_74", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_10", + "inputs": [ + "mul_74" + ], + "output": "sum_10", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_57", + "inputs": [ + "sum_10" + ], + "output": "add_57", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_26", + "inputs": [ + "add_57" + ], + "output": "rsqrt_26", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_75", + "inputs": [ + "slice_56", + "rsqrt_26" + ], + "output": "mul_75", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_4", + "inputs": [ + "mul_73", + "mul_75", + "slice_57", + "mul_71", + "sigmoid_5", + "states_11" + ], + "output": "normalized_gated_delta_rule_4", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_58", + "inputs": [ + "normalized_gated_delta_rule_4" + ], + "output": "slice_58", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_11" + }, + { + "node": "slice_59", + "inputs": [ + "normalized_gated_delta_rule_4" + ], + "output": "slice_59", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_32", + "inputs": [ + "slice_59" + ], + "output": "transpose_32", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_77", + "inputs": [ + "transpose_32", + "p_model_model_language_model_layers_5_linear_attn_norm_weight" + ], + "output": "mul_77", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_78", + "inputs": [ + "linear_40", + "mul_77" + ], + "output": "mul_78", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_43", + "inputs": [ + "mul_78", + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_43", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_59", + "inputs": [ + "add_52", + "linear_43" + ], + "output": "add_59", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_80", + "inputs": [ + "add_59", + "p_model_model_language_model_layers_5_post_attention_layernorm_weight" + ], + "output": "mul_80", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_44", + "inputs": [ + "mul_80", + "p_model_model_language_model_layers_5_mlp_gate_proj_weight", + "p_model_model_language_model_layers_5_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_44", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_45", + "inputs": [ + "mul_80", + "p_model_model_language_model_layers_5_mlp_up_proj_weight", + "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_45", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_81", + "inputs": [ + "linear_44", + "linear_45" + ], + "output": "mul_81", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_46", + "inputs": [ + "mul_81", + "p_model_model_language_model_layers_5_mlp_down_proj_weight", + "p_model_model_language_model_layers_5_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_46", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_62", + "inputs": [ + "add_59", + "linear_46" + ], + "output": "add_62", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_83", + "inputs": [ + "add_62", + "p_model_model_language_model_layers_6_input_layernorm_weight" + ], + "output": "mul_83", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_47", + "inputs": [ + "mul_83", + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_47", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_33", + "inputs": [ + "linear_47" + ], + "output": "transpose_33", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_48", + "inputs": [ + "mul_83", + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_48", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_49", + "inputs": [ + "mul_83", + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_49", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_50", + "inputs": [ + "mul_83", + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_50", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_10", + "inputs": [ + "states_12", + "transpose_33" + ], + "output": "cat_10", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_60", + "inputs": [ + "cat_10" + ], + "output": "slice_60", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_12" + }, + { + "node": "conv1d_5", + "inputs": [ + "cat_10", + "p_model_model_language_model_layers_6_linear_attn_conv1d_weight" + ], + "output": "conv1d_5", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_61", + "inputs": [ + "conv1d_5" + ], + "output": "slice_61", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_16", + "inputs": [ + "slice_61" + ], + "output": "silu_16", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_62", + "inputs": [ + "silu_16" + ], + "output": "slice_62", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_34", + "inputs": [ + "slice_62" + ], + "output": "transpose_34", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_63", + "inputs": [ + "transpose_34" + ], + "output": "slice_63", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_64", + "inputs": [ + "transpose_34" + ], + "output": "slice_64", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_65", + "inputs": [ + "transpose_34" + ], + "output": "slice_65", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_6", + "inputs": [ + "linear_49" + ], + "output": "sigmoid_6", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_5", + "inputs": [ + "p_model_model_language_model_layers_6_linear_attn_a_log" + ], + "output": "exp_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_7", + "inputs": [ + "exp_5" + ], + "output": "neg_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_65", + "inputs": [ + "linear_50", + "p_model_model_language_model_layers_6_linear_attn_dt_bias" + ], + "output": "add_65", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_5", + "inputs": [ + "add_65" + ], + "output": "softplus_5", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_84", + "inputs": [ + "neg_7", + "softplus_5" + ], + "output": "mul_84", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_85", + "inputs": [ + "slice_63" + ], + "output": "mul_85", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_11", + "inputs": [ + "mul_85" + ], + "output": "sum_11", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_66", + "inputs": [ + "sum_11" + ], + "output": "add_66", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_30", + "inputs": [ + "add_66" + ], + "output": "rsqrt_30", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_86", + "inputs": [ + "slice_63", + "rsqrt_30" + ], + "output": "mul_86", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_87", + "inputs": [ + "slice_64" + ], + "output": "mul_87", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_12", + "inputs": [ + "mul_87" + ], + "output": "sum_12", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_67", + "inputs": [ + "sum_12" + ], + "output": "add_67", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_31", + "inputs": [ + "add_67" + ], + "output": "rsqrt_31", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_88", + "inputs": [ + "slice_64", + "rsqrt_31" + ], + "output": "mul_88", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_5", + "inputs": [ + "mul_86", + "mul_88", + "slice_65", + "mul_84", + "sigmoid_6", + "states_13" + ], + "output": "normalized_gated_delta_rule_5", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_66", + "inputs": [ + "normalized_gated_delta_rule_5" + ], + "output": "slice_66", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_13" + }, + { + "node": "slice_67", + "inputs": [ + "normalized_gated_delta_rule_5" + ], + "output": "slice_67", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_35", + "inputs": [ + "slice_67" + ], + "output": "transpose_35", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_90", + "inputs": [ + "transpose_35", + "p_model_model_language_model_layers_6_linear_attn_norm_weight" + ], + "output": "mul_90", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_91", + "inputs": [ + "linear_48", + "mul_90" + ], + "output": "mul_91", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_51", + "inputs": [ + "mul_91", + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_51", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_69", + "inputs": [ + "add_62", + "linear_51" + ], + "output": "add_69", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_93", + "inputs": [ + "add_69", + "p_model_model_language_model_layers_6_post_attention_layernorm_weight" + ], + "output": "mul_93", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_52", + "inputs": [ + "mul_93", + "p_model_model_language_model_layers_6_mlp_gate_proj_weight", + "p_model_model_language_model_layers_6_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_52", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_53", + "inputs": [ + "mul_93", + "p_model_model_language_model_layers_6_mlp_up_proj_weight", + "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_53", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_94", + "inputs": [ + "linear_52", + "linear_53" + ], + "output": "mul_94", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_54", + "inputs": [ + "mul_94", + "p_model_model_language_model_layers_6_mlp_down_proj_weight", + "p_model_model_language_model_layers_6_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_54", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_72", + "inputs": [ + "add_69", + "linear_54" + ], + "output": "add_72", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_96", + "inputs": [ + "add_72", + "p_model_model_language_model_layers_7_input_layernorm_weight" + ], + "output": "mul_96", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_55", + "inputs": [ + "mul_96", + "p_model_model_language_model_layers_7_self_attn_q_proj_weight", + "p_model_model_language_model_layers_7_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_55", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/8b540ef0dbe320e8ccc513d0b118e62c6d25dff607a62b8d9de03d1a393a0a5c.wgsl" + }, + { + "node": "slice_68", + "inputs": [ + "linear_55" + ], + "output": "slice_68", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/9d3d7b777701cd320ca11358e554e1019fb041ee4d765e66071c215242906773.wgsl" + }, + { + "node": "slice_69", + "inputs": [ + "linear_55" + ], + "output": "slice_69", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/3622ffa2072335a35d6c9db5d284ed92b3fb5ed304acc2c606408063e7e21dcd.wgsl" + }, + { + "node": "mul_98", + "inputs": [ + "slice_68", + "p_model_model_language_model_layers_7_self_attn_q_norm_weight" + ], + "output": "mul_98", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl" + }, + { + "node": "transpose_36", + "inputs": [ + "mul_98" + ], + "output": "transpose_36", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/63396da046b345383e1c7b09aee551d08e4c3c45030ff5c8bb76bbc34bb2abc6.wgsl" + }, + { + "node": "linear_56", + "inputs": [ + "mul_96", + "p_model_model_language_model_layers_7_self_attn_k_proj_weight", + "p_model_model_language_model_layers_7_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_56", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl" + }, + { + "node": "mul_100", + "inputs": [ + "linear_56", + "p_model_model_language_model_layers_7_self_attn_k_norm_weight" + ], + "output": "mul_100", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl" + }, + { + "node": "transpose_37", + "inputs": [ + "mul_100" + ], + "output": "transpose_37", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl" + }, + { + "node": "linear_57", + "inputs": [ + "mul_96", + "p_model_model_language_model_layers_7_self_attn_v_proj_weight", + "p_model_model_language_model_layers_7_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_57", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl" + }, + { + "node": "transpose_38", + "inputs": [ + "linear_57" + ], + "output": "transpose_38", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl" + }, + { + "node": "slice_70", + "inputs": [ + "transpose_36" + ], + "output": "slice_70", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f2ee90b3d949141c23610237d27d44a7b833aacae8cdbc2e42a595d5eb7b5537.wgsl" + }, + { + "node": "slice_71", + "inputs": [ + "transpose_36" + ], + "output": "slice_71", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/8d62df48089497dd19fea48680730210decda42b63f69ecdf9688fc4a43efc58.wgsl" + }, + { + "node": "slice_72", + "inputs": [ + "transpose_37" + ], + "output": "slice_72", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6caa35fcab56c639d1859e1b090a7e5bf016dd7f33ecd2373c24ab625de16e09.wgsl" + }, + { + "node": "slice_73", + "inputs": [ + "transpose_37" + ], + "output": "slice_73", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/2182574f59b7d9e38d74b30540c5781827e7da2de1ba6b00327af0b4cb42379d.wgsl" + }, + { + "node": "mul_101", + "inputs": [ + "slice_70", + "mul" + ], + "output": "mul_101", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e3c6a64e01b6ea79adf125ff00a6dcf00f0b65aee0f4a3d14fdbc20e5dfeb22d.wgsl" + }, + { + "node": "slice_74", + "inputs": [ + "slice_70" + ], + "output": "slice_74", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/b5984ff6eb9e56ef0f9930f775c1dc3c203eec15924e5fea4c2525b2ce180857.wgsl" + }, + { + "node": "slice_75", + "inputs": [ + "slice_70" + ], + "output": "slice_75", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/8256ea6cdc3500306470c0bb2dcdb44ebdd454dafa218ad6a94dd953b9bbe125.wgsl" + }, + { + "node": "neg_8", + "inputs": [ + "slice_75" + ], + "output": "neg_8", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/adf1e58a124e1301625024ae142dca5fb1dd50799fa94b52eaeffb82ea99852c.wgsl" + }, + { + "node": "cat_11", + "inputs": [ + "neg_8", + "slice_74" + ], + "output": "cat_11", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl" + }, + { + "node": "mul_102", + "inputs": [ + "cat_11", + "mul_1" + ], + "output": "mul_102", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e3c6a64e01b6ea79adf125ff00a6dcf00f0b65aee0f4a3d14fdbc20e5dfeb22d.wgsl" + }, + { + "node": "add_79", + "inputs": [ + "mul_101", + "mul_102" + ], + "output": "add_79", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl" + }, + { + "node": "mul_103", + "inputs": [ + "slice_72", + "mul" + ], + "output": "mul_103", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl" + }, + { + "node": "slice_76", + "inputs": [ + "slice_72" + ], + "output": "slice_76", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/58077eb80a5fa253f535c97c82dc07edd48455a17cc52583aeee41c7d1bbd0b3.wgsl" + }, + { + "node": "slice_77", + "inputs": [ + "slice_72" + ], + "output": "slice_77", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3830061b0fc54c85d3d8a783b927c44484b1a3b68b52b8d8a1b980f7553deebf.wgsl" + }, + { + "node": "neg_9", + "inputs": [ + "slice_77" + ], + "output": "neg_9", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl" + }, + { + "node": "cat_12", + "inputs": [ + "neg_9", + "slice_76" + ], + "output": "cat_12", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/bda7ebd3bcf25c31938fca1a582e82b972ab742f85f826f085417b0e6f651e60.wgsl" + }, + { + "node": "mul_104", + "inputs": [ + "cat_12", + "mul_1" + ], + "output": "mul_104", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl" + }, + { + "node": "add_80", + "inputs": [ + "mul_103", + "mul_104" + ], + "output": "add_80", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl" + }, + { + "node": "cat_13", + "inputs": [ + "add_79", + "slice_71" + ], + "output": "cat_13", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl" + }, + { + "node": "cat_14", + "inputs": [ + "add_80", + "slice_73" + ], + "output": "cat_14", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl" + }, + { + "node": "cache_update_2", + "inputs": [ + "cat_14", + "position_ids" + ], + "output": "cache_update_2", + "workgroups": [ + 128, + 1, + 1 + ], + "inPlaceOf": "states_14", + "kernel": "kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl" + }, + { + "node": "cache_update_3", + "inputs": [ + "transpose_38", + "position_ids" + ], + "output": "cache_update_3", + "workgroups": [ + 128, + 1, + 1 + ], + "inPlaceOf": "states_15", + "kernel": "kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl" + }, + { + "node": "matmul_4", + "inputs": [ + "cat_13", + "cache_update_2", + "cache_update_3", + "attention_mask" + ], + "output": "matmul_4", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d9ab25f0ef4248028d7dab1e782d0756dfc664cc90794d13644bd91106541608.wgsl" + }, + { + "node": "transpose_40", + "inputs": [ + "matmul_4" + ], + "output": "transpose_40", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/84f24861effe9e2f3b05f98fe6d4e16f4f8204be40a7eeece0668e57060ecc7b.wgsl" + }, + { + "node": "sigmoid_7", + "inputs": [ + "slice_69" + ], + "output": "sigmoid_7", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/816861e23c92e156c1ac01f21e4f6dd1cbfc14978b1be8a5994fb45b0f845cb8.wgsl" + }, + { + "node": "mul_106", + "inputs": [ + "transpose_40", + "sigmoid_7" + ], + "output": "mul_106", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/aa227ae49fab3f75d86e02e6f390be785eabfeafe70e77a5c7eb4ad51895e382.wgsl" + }, + { + "node": "linear_58", + "inputs": [ + "mul_106", + "p_model_model_language_model_layers_7_self_attn_o_proj_weight", + "p_model_model_language_model_layers_7_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_58", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_82", + "inputs": [ + "add_72", + "linear_58" + ], + "output": "add_82", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_108", + "inputs": [ + "add_82", + "p_model_model_language_model_layers_7_post_attention_layernorm_weight" + ], + "output": "mul_108", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_59", + "inputs": [ + "mul_108", + "p_model_model_language_model_layers_7_mlp_gate_proj_weight", + "p_model_model_language_model_layers_7_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_59", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_60", + "inputs": [ + "mul_108", + "p_model_model_language_model_layers_7_mlp_up_proj_weight", + "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_60", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_109", + "inputs": [ + "linear_59", + "linear_60" + ], + "output": "mul_109", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_61", + "inputs": [ + "mul_109", + "p_model_model_language_model_layers_7_mlp_down_proj_weight", + "p_model_model_language_model_layers_7_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_61", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_85", + "inputs": [ + "add_82", + "linear_61" + ], + "output": "add_85", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_111", + "inputs": [ + "add_85", + "p_model_model_language_model_layers_8_input_layernorm_weight" + ], + "output": "mul_111", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_62", + "inputs": [ + "mul_111", + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_62", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_41", + "inputs": [ + "linear_62" + ], + "output": "transpose_41", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_63", + "inputs": [ + "mul_111", + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_63", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_64", + "inputs": [ + "mul_111", + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_64", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_65", + "inputs": [ + "mul_111", + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_65", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_15", + "inputs": [ + "states_16", + "transpose_41" + ], + "output": "cat_15", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_78", + "inputs": [ + "cat_15" + ], + "output": "slice_78", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_16" + }, + { + "node": "conv1d_6", + "inputs": [ + "cat_15", + "p_model_model_language_model_layers_8_linear_attn_conv1d_weight" + ], + "output": "conv1d_6", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_79", + "inputs": [ + "conv1d_6" + ], + "output": "slice_79", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_20", + "inputs": [ + "slice_79" + ], + "output": "silu_20", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_80", + "inputs": [ + "silu_20" + ], + "output": "slice_80", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_42", + "inputs": [ + "slice_80" + ], + "output": "transpose_42", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_81", + "inputs": [ + "transpose_42" + ], + "output": "slice_81", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_82", + "inputs": [ + "transpose_42" + ], + "output": "slice_82", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_83", + "inputs": [ + "transpose_42" + ], + "output": "slice_83", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_8", + "inputs": [ + "linear_64" + ], + "output": "sigmoid_8", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_6", + "inputs": [ + "p_model_model_language_model_layers_8_linear_attn_a_log" + ], + "output": "exp_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_10", + "inputs": [ + "exp_6" + ], + "output": "neg_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_88", + "inputs": [ + "linear_65", + "p_model_model_language_model_layers_8_linear_attn_dt_bias" + ], + "output": "add_88", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_6", + "inputs": [ + "add_88" + ], + "output": "softplus_6", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_112", + "inputs": [ + "neg_10", + "softplus_6" + ], + "output": "mul_112", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_113", + "inputs": [ + "slice_81" + ], + "output": "mul_113", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_13", + "inputs": [ + "mul_113" + ], + "output": "sum_13", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_89", + "inputs": [ + "sum_13" + ], + "output": "add_89", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_39", + "inputs": [ + "add_89" + ], + "output": "rsqrt_39", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_114", + "inputs": [ + "slice_81", + "rsqrt_39" + ], + "output": "mul_114", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_115", + "inputs": [ + "slice_82" + ], + "output": "mul_115", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_14", + "inputs": [ + "mul_115" + ], + "output": "sum_14", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_90", + "inputs": [ + "sum_14" + ], + "output": "add_90", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_40", + "inputs": [ + "add_90" + ], + "output": "rsqrt_40", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_116", + "inputs": [ + "slice_82", + "rsqrt_40" + ], + "output": "mul_116", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_6", + "inputs": [ + "mul_114", + "mul_116", + "slice_83", + "mul_112", + "sigmoid_8", + "states_17" + ], + "output": "normalized_gated_delta_rule_6", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_84", + "inputs": [ + "normalized_gated_delta_rule_6" + ], + "output": "slice_84", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_17" + }, + { + "node": "slice_85", + "inputs": [ + "normalized_gated_delta_rule_6" + ], + "output": "slice_85", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_43", + "inputs": [ + "slice_85" + ], + "output": "transpose_43", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_118", + "inputs": [ + "transpose_43", + "p_model_model_language_model_layers_8_linear_attn_norm_weight" + ], + "output": "mul_118", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_119", + "inputs": [ + "linear_63", + "mul_118" + ], + "output": "mul_119", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_66", + "inputs": [ + "mul_119", + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_66", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_92", + "inputs": [ + "add_85", + "linear_66" + ], + "output": "add_92", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_121", + "inputs": [ + "add_92", + "p_model_model_language_model_layers_8_post_attention_layernorm_weight" + ], + "output": "mul_121", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_67", + "inputs": [ + "mul_121", + "p_model_model_language_model_layers_8_mlp_gate_proj_weight", + "p_model_model_language_model_layers_8_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_67", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_68", + "inputs": [ + "mul_121", + "p_model_model_language_model_layers_8_mlp_up_proj_weight", + "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_68", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_122", + "inputs": [ + "linear_67", + "linear_68" + ], + "output": "mul_122", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_69", + "inputs": [ + "mul_122", + "p_model_model_language_model_layers_8_mlp_down_proj_weight", + "p_model_model_language_model_layers_8_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_69", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_95", + "inputs": [ + "add_92", + "linear_69" + ], + "output": "add_95", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_124", + "inputs": [ + "add_95", + "p_model_model_language_model_layers_9_input_layernorm_weight" + ], + "output": "mul_124", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_70", + "inputs": [ + "mul_124", + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_70", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_44", + "inputs": [ + "linear_70" + ], + "output": "transpose_44", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_71", + "inputs": [ + "mul_124", + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_71", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_72", + "inputs": [ + "mul_124", + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_72", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_73", + "inputs": [ + "mul_124", + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_73", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_16", + "inputs": [ + "states_18", + "transpose_44" + ], + "output": "cat_16", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_86", + "inputs": [ + "cat_16" + ], + "output": "slice_86", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_18" + }, + { + "node": "conv1d_7", + "inputs": [ + "cat_16", + "p_model_model_language_model_layers_9_linear_attn_conv1d_weight" + ], + "output": "conv1d_7", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_87", + "inputs": [ + "conv1d_7" + ], + "output": "slice_87", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_23", + "inputs": [ + "slice_87" + ], + "output": "silu_23", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_88", + "inputs": [ + "silu_23" + ], + "output": "slice_88", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_45", + "inputs": [ + "slice_88" + ], + "output": "transpose_45", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_89", + "inputs": [ + "transpose_45" + ], + "output": "slice_89", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_90", + "inputs": [ + "transpose_45" + ], + "output": "slice_90", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_91", + "inputs": [ + "transpose_45" + ], + "output": "slice_91", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_9", + "inputs": [ + "linear_72" + ], + "output": "sigmoid_9", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_7", + "inputs": [ + "p_model_model_language_model_layers_9_linear_attn_a_log" + ], + "output": "exp_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_11", + "inputs": [ + "exp_7" + ], + "output": "neg_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_98", + "inputs": [ + "linear_73", + "p_model_model_language_model_layers_9_linear_attn_dt_bias" + ], + "output": "add_98", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_7", + "inputs": [ + "add_98" + ], + "output": "softplus_7", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_125", + "inputs": [ + "neg_11", + "softplus_7" + ], + "output": "mul_125", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_126", + "inputs": [ + "slice_89" + ], + "output": "mul_126", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_15", + "inputs": [ + "mul_126" + ], + "output": "sum_15", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_99", + "inputs": [ + "sum_15" + ], + "output": "add_99", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_44", + "inputs": [ + "add_99" + ], + "output": "rsqrt_44", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_127", + "inputs": [ + "slice_89", + "rsqrt_44" + ], + "output": "mul_127", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_128", + "inputs": [ + "slice_90" + ], + "output": "mul_128", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_16", + "inputs": [ + "mul_128" + ], + "output": "sum_16", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_100", + "inputs": [ + "sum_16" + ], + "output": "add_100", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_45", + "inputs": [ + "add_100" + ], + "output": "rsqrt_45", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_129", + "inputs": [ + "slice_90", + "rsqrt_45" + ], + "output": "mul_129", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_7", + "inputs": [ + "mul_127", + "mul_129", + "slice_91", + "mul_125", + "sigmoid_9", + "states_19" + ], + "output": "normalized_gated_delta_rule_7", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_92", + "inputs": [ + "normalized_gated_delta_rule_7" + ], + "output": "slice_92", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_19" + }, + { + "node": "slice_93", + "inputs": [ + "normalized_gated_delta_rule_7" + ], + "output": "slice_93", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_46", + "inputs": [ + "slice_93" + ], + "output": "transpose_46", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_131", + "inputs": [ + "transpose_46", + "p_model_model_language_model_layers_9_linear_attn_norm_weight" + ], + "output": "mul_131", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_132", + "inputs": [ + "linear_71", + "mul_131" + ], + "output": "mul_132", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_74", + "inputs": [ + "mul_132", + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_74", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_102", + "inputs": [ + "add_95", + "linear_74" + ], + "output": "add_102", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_134", + "inputs": [ + "add_102", + "p_model_model_language_model_layers_9_post_attention_layernorm_weight" + ], + "output": "mul_134", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_75", + "inputs": [ + "mul_134", + "p_model_model_language_model_layers_9_mlp_gate_proj_weight", + "p_model_model_language_model_layers_9_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_75", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_76", + "inputs": [ + "mul_134", + "p_model_model_language_model_layers_9_mlp_up_proj_weight", + "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_76", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_135", + "inputs": [ + "linear_75", + "linear_76" + ], + "output": "mul_135", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_77", + "inputs": [ + "mul_135", + "p_model_model_language_model_layers_9_mlp_down_proj_weight", + "p_model_model_language_model_layers_9_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_77", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_105", + "inputs": [ + "add_102", + "linear_77" + ], + "output": "add_105", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_137", + "inputs": [ + "add_105", + "p_model_model_language_model_layers_10_input_layernorm_weight" + ], + "output": "mul_137", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_78", + "inputs": [ + "mul_137", + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_78", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_47", + "inputs": [ + "linear_78" + ], + "output": "transpose_47", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_79", + "inputs": [ + "mul_137", + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_79", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_80", + "inputs": [ + "mul_137", + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_80", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_81", + "inputs": [ + "mul_137", + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_81", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_17", + "inputs": [ + "states_20", + "transpose_47" + ], + "output": "cat_17", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_94", + "inputs": [ + "cat_17" + ], + "output": "slice_94", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_20" + }, + { + "node": "conv1d_8", + "inputs": [ + "cat_17", + "p_model_model_language_model_layers_10_linear_attn_conv1d_weight" + ], + "output": "conv1d_8", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_95", + "inputs": [ + "conv1d_8" + ], + "output": "slice_95", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_26", + "inputs": [ + "slice_95" + ], + "output": "silu_26", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_96", + "inputs": [ + "silu_26" + ], + "output": "slice_96", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_48", + "inputs": [ + "slice_96" + ], + "output": "transpose_48", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_97", + "inputs": [ + "transpose_48" + ], + "output": "slice_97", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_98", + "inputs": [ + "transpose_48" + ], + "output": "slice_98", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_99", + "inputs": [ + "transpose_48" + ], + "output": "slice_99", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_10", + "inputs": [ + "linear_80" + ], + "output": "sigmoid_10", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_8", + "inputs": [ + "p_model_model_language_model_layers_10_linear_attn_a_log" + ], + "output": "exp_8", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_12", + "inputs": [ + "exp_8" + ], + "output": "neg_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_108", + "inputs": [ + "linear_81", + "p_model_model_language_model_layers_10_linear_attn_dt_bias" + ], + "output": "add_108", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_8", + "inputs": [ + "add_108" + ], + "output": "softplus_8", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_138", + "inputs": [ + "neg_12", + "softplus_8" + ], + "output": "mul_138", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_139", + "inputs": [ + "slice_97" + ], + "output": "mul_139", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_17", + "inputs": [ + "mul_139" + ], + "output": "sum_17", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_109", + "inputs": [ + "sum_17" + ], + "output": "add_109", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_49", + "inputs": [ + "add_109" + ], + "output": "rsqrt_49", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_140", + "inputs": [ + "slice_97", + "rsqrt_49" + ], + "output": "mul_140", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_141", + "inputs": [ + "slice_98" + ], + "output": "mul_141", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_18", + "inputs": [ + "mul_141" + ], + "output": "sum_18", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_110", + "inputs": [ + "sum_18" + ], + "output": "add_110", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_50", + "inputs": [ + "add_110" + ], + "output": "rsqrt_50", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_142", + "inputs": [ + "slice_98", + "rsqrt_50" + ], + "output": "mul_142", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_8", + "inputs": [ + "mul_140", + "mul_142", + "slice_99", + "mul_138", + "sigmoid_10", + "states_21" + ], + "output": "normalized_gated_delta_rule_8", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_100", + "inputs": [ + "normalized_gated_delta_rule_8" + ], + "output": "slice_100", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_21" + }, + { + "node": "slice_101", + "inputs": [ + "normalized_gated_delta_rule_8" + ], + "output": "slice_101", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_49", + "inputs": [ + "slice_101" + ], + "output": "transpose_49", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_144", + "inputs": [ + "transpose_49", + "p_model_model_language_model_layers_10_linear_attn_norm_weight" + ], + "output": "mul_144", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_145", + "inputs": [ + "linear_79", + "mul_144" + ], + "output": "mul_145", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_82", + "inputs": [ + "mul_145", + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_82", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_112", + "inputs": [ + "add_105", + "linear_82" + ], + "output": "add_112", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_147", + "inputs": [ + "add_112", + "p_model_model_language_model_layers_10_post_attention_layernorm_weight" + ], + "output": "mul_147", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_83", + "inputs": [ + "mul_147", + "p_model_model_language_model_layers_10_mlp_gate_proj_weight", + "p_model_model_language_model_layers_10_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_83", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_84", + "inputs": [ + "mul_147", + "p_model_model_language_model_layers_10_mlp_up_proj_weight", + "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_84", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_148", + "inputs": [ + "linear_83", + "linear_84" + ], + "output": "mul_148", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_85", + "inputs": [ + "mul_148", + "p_model_model_language_model_layers_10_mlp_down_proj_weight", + "p_model_model_language_model_layers_10_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_85", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_115", + "inputs": [ + "add_112", + "linear_85" + ], + "output": "add_115", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_150", + "inputs": [ + "add_115", + "p_model_model_language_model_layers_11_input_layernorm_weight" + ], + "output": "mul_150", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_86", + "inputs": [ + "mul_150", + "p_model_model_language_model_layers_11_self_attn_q_proj_weight", + "p_model_model_language_model_layers_11_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_86", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/8b540ef0dbe320e8ccc513d0b118e62c6d25dff607a62b8d9de03d1a393a0a5c.wgsl" + }, + { + "node": "slice_102", + "inputs": [ + "linear_86" + ], + "output": "slice_102", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/9d3d7b777701cd320ca11358e554e1019fb041ee4d765e66071c215242906773.wgsl" + }, + { + "node": "slice_103", + "inputs": [ + "linear_86" + ], + "output": "slice_103", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/3622ffa2072335a35d6c9db5d284ed92b3fb5ed304acc2c606408063e7e21dcd.wgsl" + }, + { + "node": "mul_152", + "inputs": [ + "slice_102", + "p_model_model_language_model_layers_11_self_attn_q_norm_weight" + ], + "output": "mul_152", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl" + }, + { + "node": "transpose_50", + "inputs": [ + "mul_152" + ], + "output": "transpose_50", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/63396da046b345383e1c7b09aee551d08e4c3c45030ff5c8bb76bbc34bb2abc6.wgsl" + }, + { + "node": "linear_87", + "inputs": [ + "mul_150", + "p_model_model_language_model_layers_11_self_attn_k_proj_weight", + "p_model_model_language_model_layers_11_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_87", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl" + }, + { + "node": "mul_154", + "inputs": [ + "linear_87", + "p_model_model_language_model_layers_11_self_attn_k_norm_weight" + ], + "output": "mul_154", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl" + }, + { + "node": "transpose_51", + "inputs": [ + "mul_154" + ], + "output": "transpose_51", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl" + }, + { + "node": "linear_88", + "inputs": [ + "mul_150", + "p_model_model_language_model_layers_11_self_attn_v_proj_weight", + "p_model_model_language_model_layers_11_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_88", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl" + }, + { + "node": "transpose_52", + "inputs": [ + "linear_88" + ], + "output": "transpose_52", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl" + }, + { + "node": "slice_104", + "inputs": [ + "transpose_50" + ], + "output": "slice_104", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f2ee90b3d949141c23610237d27d44a7b833aacae8cdbc2e42a595d5eb7b5537.wgsl" + }, + { + "node": "slice_105", + "inputs": [ + "transpose_50" + ], + "output": "slice_105", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/8d62df48089497dd19fea48680730210decda42b63f69ecdf9688fc4a43efc58.wgsl" + }, + { + "node": "slice_106", + "inputs": [ + "transpose_51" + ], + "output": "slice_106", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6caa35fcab56c639d1859e1b090a7e5bf016dd7f33ecd2373c24ab625de16e09.wgsl" + }, + { + "node": "slice_107", + "inputs": [ + "transpose_51" + ], + "output": "slice_107", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/2182574f59b7d9e38d74b30540c5781827e7da2de1ba6b00327af0b4cb42379d.wgsl" + }, + { + "node": "mul_155", + "inputs": [ + "slice_104", + "mul" + ], + "output": "mul_155", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e3c6a64e01b6ea79adf125ff00a6dcf00f0b65aee0f4a3d14fdbc20e5dfeb22d.wgsl" + }, + { + "node": "slice_108", + "inputs": [ + "slice_104" + ], + "output": "slice_108", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/b5984ff6eb9e56ef0f9930f775c1dc3c203eec15924e5fea4c2525b2ce180857.wgsl" + }, + { + "node": "slice_109", + "inputs": [ + "slice_104" + ], + "output": "slice_109", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/8256ea6cdc3500306470c0bb2dcdb44ebdd454dafa218ad6a94dd953b9bbe125.wgsl" + }, + { + "node": "neg_13", + "inputs": [ + "slice_109" + ], + "output": "neg_13", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/adf1e58a124e1301625024ae142dca5fb1dd50799fa94b52eaeffb82ea99852c.wgsl" + }, + { + "node": "cat_18", + "inputs": [ + "neg_13", + "slice_108" + ], + "output": "cat_18", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl" + }, + { + "node": "mul_156", + "inputs": [ + "cat_18", + "mul_1" + ], + "output": "mul_156", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e3c6a64e01b6ea79adf125ff00a6dcf00f0b65aee0f4a3d14fdbc20e5dfeb22d.wgsl" + }, + { + "node": "add_122", + "inputs": [ + "mul_155", + "mul_156" + ], + "output": "add_122", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl" + }, + { + "node": "mul_157", + "inputs": [ + "slice_106", + "mul" + ], + "output": "mul_157", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl" + }, + { + "node": "slice_110", + "inputs": [ + "slice_106" + ], + "output": "slice_110", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/58077eb80a5fa253f535c97c82dc07edd48455a17cc52583aeee41c7d1bbd0b3.wgsl" + }, + { + "node": "slice_111", + "inputs": [ + "slice_106" + ], + "output": "slice_111", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3830061b0fc54c85d3d8a783b927c44484b1a3b68b52b8d8a1b980f7553deebf.wgsl" + }, + { + "node": "neg_14", + "inputs": [ + "slice_111" + ], + "output": "neg_14", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl" + }, + { + "node": "cat_19", + "inputs": [ + "neg_14", + "slice_110" + ], + "output": "cat_19", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/bda7ebd3bcf25c31938fca1a582e82b972ab742f85f826f085417b0e6f651e60.wgsl" + }, + { + "node": "mul_158", + "inputs": [ + "cat_19", + "mul_1" + ], + "output": "mul_158", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl" + }, + { + "node": "add_123", + "inputs": [ + "mul_157", + "mul_158" + ], + "output": "add_123", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl" + }, + { + "node": "cat_20", + "inputs": [ + "add_122", + "slice_105" + ], + "output": "cat_20", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl" + }, + { + "node": "cat_21", + "inputs": [ + "add_123", + "slice_107" + ], + "output": "cat_21", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl" + }, + { + "node": "cache_update_4", + "inputs": [ + "cat_21", + "position_ids" + ], + "output": "cache_update_4", + "workgroups": [ + 128, + 1, + 1 + ], + "inPlaceOf": "states_22", + "kernel": "kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl" + }, + { + "node": "cache_update_5", + "inputs": [ + "transpose_52", + "position_ids" + ], + "output": "cache_update_5", + "workgroups": [ + 128, + 1, + 1 + ], + "inPlaceOf": "states_23", + "kernel": "kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl" + }, + { + "node": "matmul_6", + "inputs": [ + "cat_20", + "cache_update_4", + "cache_update_5", + "attention_mask" + ], + "output": "matmul_6", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d9ab25f0ef4248028d7dab1e782d0756dfc664cc90794d13644bd91106541608.wgsl" + }, + { + "node": "transpose_54", + "inputs": [ + "matmul_6" + ], + "output": "transpose_54", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/84f24861effe9e2f3b05f98fe6d4e16f4f8204be40a7eeece0668e57060ecc7b.wgsl" + }, + { + "node": "sigmoid_11", + "inputs": [ + "slice_103" + ], + "output": "sigmoid_11", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/816861e23c92e156c1ac01f21e4f6dd1cbfc14978b1be8a5994fb45b0f845cb8.wgsl" + }, + { + "node": "mul_160", + "inputs": [ + "transpose_54", + "sigmoid_11" + ], + "output": "mul_160", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/aa227ae49fab3f75d86e02e6f390be785eabfeafe70e77a5c7eb4ad51895e382.wgsl" + }, + { + "node": "linear_89", + "inputs": [ + "mul_160", + "p_model_model_language_model_layers_11_self_attn_o_proj_weight", + "p_model_model_language_model_layers_11_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_89", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_125", + "inputs": [ + "add_115", + "linear_89" + ], + "output": "add_125", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_162", + "inputs": [ + "add_125", + "p_model_model_language_model_layers_11_post_attention_layernorm_weight" + ], + "output": "mul_162", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_90", + "inputs": [ + "mul_162", + "p_model_model_language_model_layers_11_mlp_gate_proj_weight", + "p_model_model_language_model_layers_11_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_90", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_91", + "inputs": [ + "mul_162", + "p_model_model_language_model_layers_11_mlp_up_proj_weight", + "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_91", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_163", + "inputs": [ + "linear_90", + "linear_91" + ], + "output": "mul_163", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_92", + "inputs": [ + "mul_163", + "p_model_model_language_model_layers_11_mlp_down_proj_weight", + "p_model_model_language_model_layers_11_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_92", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_128", + "inputs": [ + "add_125", + "linear_92" + ], + "output": "add_128", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_165", + "inputs": [ + "add_128", + "p_model_model_language_model_layers_12_input_layernorm_weight" + ], + "output": "mul_165", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_93", + "inputs": [ + "mul_165", + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_93", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_55", + "inputs": [ + "linear_93" + ], + "output": "transpose_55", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_94", + "inputs": [ + "mul_165", + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_94", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_95", + "inputs": [ + "mul_165", + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_95", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_96", + "inputs": [ + "mul_165", + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_96", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_22", + "inputs": [ + "states_24", + "transpose_55" + ], + "output": "cat_22", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_112", + "inputs": [ + "cat_22" + ], + "output": "slice_112", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_24" + }, + { + "node": "conv1d_9", + "inputs": [ + "cat_22", + "p_model_model_language_model_layers_12_linear_attn_conv1d_weight" + ], + "output": "conv1d_9", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_113", + "inputs": [ + "conv1d_9" + ], + "output": "slice_113", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_30", + "inputs": [ + "slice_113" + ], + "output": "silu_30", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_114", + "inputs": [ + "silu_30" + ], + "output": "slice_114", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_56", + "inputs": [ + "slice_114" + ], + "output": "transpose_56", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_115", + "inputs": [ + "transpose_56" + ], + "output": "slice_115", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_116", + "inputs": [ + "transpose_56" + ], + "output": "slice_116", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_117", + "inputs": [ + "transpose_56" + ], + "output": "slice_117", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_12", + "inputs": [ + "linear_95" + ], + "output": "sigmoid_12", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_9", + "inputs": [ + "p_model_model_language_model_layers_12_linear_attn_a_log" + ], + "output": "exp_9", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_15", + "inputs": [ + "exp_9" + ], + "output": "neg_15", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_131", + "inputs": [ + "linear_96", + "p_model_model_language_model_layers_12_linear_attn_dt_bias" + ], + "output": "add_131", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_9", + "inputs": [ + "add_131" + ], + "output": "softplus_9", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_166", + "inputs": [ + "neg_15", + "softplus_9" + ], + "output": "mul_166", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_167", + "inputs": [ + "slice_115" + ], + "output": "mul_167", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_19", + "inputs": [ + "mul_167" + ], + "output": "sum_19", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_132", + "inputs": [ + "sum_19" + ], + "output": "add_132", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_58", + "inputs": [ + "add_132" + ], + "output": "rsqrt_58", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_168", + "inputs": [ + "slice_115", + "rsqrt_58" + ], + "output": "mul_168", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_169", + "inputs": [ + "slice_116" + ], + "output": "mul_169", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_20", + "inputs": [ + "mul_169" + ], + "output": "sum_20", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_133", + "inputs": [ + "sum_20" + ], + "output": "add_133", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_59", + "inputs": [ + "add_133" + ], + "output": "rsqrt_59", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_170", + "inputs": [ + "slice_116", + "rsqrt_59" + ], + "output": "mul_170", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_9", + "inputs": [ + "mul_168", + "mul_170", + "slice_117", + "mul_166", + "sigmoid_12", + "states_25" + ], + "output": "normalized_gated_delta_rule_9", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_118", + "inputs": [ + "normalized_gated_delta_rule_9" + ], + "output": "slice_118", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_25" + }, + { + "node": "slice_119", + "inputs": [ + "normalized_gated_delta_rule_9" + ], + "output": "slice_119", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_57", + "inputs": [ + "slice_119" + ], + "output": "transpose_57", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_172", + "inputs": [ + "transpose_57", + "p_model_model_language_model_layers_12_linear_attn_norm_weight" + ], + "output": "mul_172", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_173", + "inputs": [ + "linear_94", + "mul_172" + ], + "output": "mul_173", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_97", + "inputs": [ + "mul_173", + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_97", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_135", + "inputs": [ + "add_128", + "linear_97" + ], + "output": "add_135", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_175", + "inputs": [ + "add_135", + "p_model_model_language_model_layers_12_post_attention_layernorm_weight" + ], + "output": "mul_175", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_98", + "inputs": [ + "mul_175", + "p_model_model_language_model_layers_12_mlp_gate_proj_weight", + "p_model_model_language_model_layers_12_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_98", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_99", + "inputs": [ + "mul_175", + "p_model_model_language_model_layers_12_mlp_up_proj_weight", + "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_99", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_176", + "inputs": [ + "linear_98", + "linear_99" + ], + "output": "mul_176", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_100", + "inputs": [ + "mul_176", + "p_model_model_language_model_layers_12_mlp_down_proj_weight", + "p_model_model_language_model_layers_12_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_100", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_138", + "inputs": [ + "add_135", + "linear_100" + ], + "output": "add_138", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_178", + "inputs": [ + "add_138", + "p_model_model_language_model_layers_13_input_layernorm_weight" + ], + "output": "mul_178", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_101", + "inputs": [ + "mul_178", + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_101", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_58", + "inputs": [ + "linear_101" + ], + "output": "transpose_58", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_102", + "inputs": [ + "mul_178", + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_102", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_103", + "inputs": [ + "mul_178", + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_103", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_104", + "inputs": [ + "mul_178", + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_104", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_23", + "inputs": [ + "states_26", + "transpose_58" + ], + "output": "cat_23", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_120", + "inputs": [ + "cat_23" + ], + "output": "slice_120", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_26" + }, + { + "node": "conv1d_10", + "inputs": [ + "cat_23", + "p_model_model_language_model_layers_13_linear_attn_conv1d_weight" + ], + "output": "conv1d_10", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_121", + "inputs": [ + "conv1d_10" + ], + "output": "slice_121", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_33", + "inputs": [ + "slice_121" + ], + "output": "silu_33", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_122", + "inputs": [ + "silu_33" + ], + "output": "slice_122", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_59", + "inputs": [ + "slice_122" + ], + "output": "transpose_59", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_123", + "inputs": [ + "transpose_59" + ], + "output": "slice_123", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_124", + "inputs": [ + "transpose_59" + ], + "output": "slice_124", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_125", + "inputs": [ + "transpose_59" + ], + "output": "slice_125", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_13", + "inputs": [ + "linear_103" + ], + "output": "sigmoid_13", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_10", + "inputs": [ + "p_model_model_language_model_layers_13_linear_attn_a_log" + ], + "output": "exp_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_16", + "inputs": [ + "exp_10" + ], + "output": "neg_16", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_141", + "inputs": [ + "linear_104", + "p_model_model_language_model_layers_13_linear_attn_dt_bias" + ], + "output": "add_141", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_10", + "inputs": [ + "add_141" + ], + "output": "softplus_10", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_179", + "inputs": [ + "neg_16", + "softplus_10" + ], + "output": "mul_179", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_180", + "inputs": [ + "slice_123" + ], + "output": "mul_180", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_21", + "inputs": [ + "mul_180" + ], + "output": "sum_21", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_142", + "inputs": [ + "sum_21" + ], + "output": "add_142", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_63", + "inputs": [ + "add_142" + ], + "output": "rsqrt_63", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_181", + "inputs": [ + "slice_123", + "rsqrt_63" + ], + "output": "mul_181", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_182", + "inputs": [ + "slice_124" + ], + "output": "mul_182", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_22", + "inputs": [ + "mul_182" + ], + "output": "sum_22", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_143", + "inputs": [ + "sum_22" + ], + "output": "add_143", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_64", + "inputs": [ + "add_143" + ], + "output": "rsqrt_64", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_183", + "inputs": [ + "slice_124", + "rsqrt_64" + ], + "output": "mul_183", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_10", + "inputs": [ + "mul_181", + "mul_183", + "slice_125", + "mul_179", + "sigmoid_13", + "states_27" + ], + "output": "normalized_gated_delta_rule_10", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_126", + "inputs": [ + "normalized_gated_delta_rule_10" + ], + "output": "slice_126", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_27" + }, + { + "node": "slice_127", + "inputs": [ + "normalized_gated_delta_rule_10" + ], + "output": "slice_127", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_60", + "inputs": [ + "slice_127" + ], + "output": "transpose_60", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_185", + "inputs": [ + "transpose_60", + "p_model_model_language_model_layers_13_linear_attn_norm_weight" + ], + "output": "mul_185", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_186", + "inputs": [ + "linear_102", + "mul_185" + ], + "output": "mul_186", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_105", + "inputs": [ + "mul_186", + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_105", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_145", + "inputs": [ + "add_138", + "linear_105" + ], + "output": "add_145", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_188", + "inputs": [ + "add_145", + "p_model_model_language_model_layers_13_post_attention_layernorm_weight" + ], + "output": "mul_188", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_106", + "inputs": [ + "mul_188", + "p_model_model_language_model_layers_13_mlp_gate_proj_weight", + "p_model_model_language_model_layers_13_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_106", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_107", + "inputs": [ + "mul_188", + "p_model_model_language_model_layers_13_mlp_up_proj_weight", + "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_107", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_189", + "inputs": [ + "linear_106", + "linear_107" + ], + "output": "mul_189", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_108", + "inputs": [ + "mul_189", + "p_model_model_language_model_layers_13_mlp_down_proj_weight", + "p_model_model_language_model_layers_13_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_108", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_148", + "inputs": [ + "add_145", + "linear_108" + ], + "output": "add_148", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_191", + "inputs": [ + "add_148", + "p_model_model_language_model_layers_14_input_layernorm_weight" + ], + "output": "mul_191", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_109", + "inputs": [ + "mul_191", + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_109", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_61", + "inputs": [ + "linear_109" + ], + "output": "transpose_61", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_110", + "inputs": [ + "mul_191", + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_110", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_111", + "inputs": [ + "mul_191", + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_111", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_112", + "inputs": [ + "mul_191", + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_112", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_24", + "inputs": [ + "states_28", + "transpose_61" + ], + "output": "cat_24", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_128", + "inputs": [ + "cat_24" + ], + "output": "slice_128", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_28" + }, + { + "node": "conv1d_11", + "inputs": [ + "cat_24", + "p_model_model_language_model_layers_14_linear_attn_conv1d_weight" + ], + "output": "conv1d_11", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_129", + "inputs": [ + "conv1d_11" + ], + "output": "slice_129", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_36", + "inputs": [ + "slice_129" + ], + "output": "silu_36", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_130", + "inputs": [ + "silu_36" + ], + "output": "slice_130", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_62", + "inputs": [ + "slice_130" + ], + "output": "transpose_62", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_131", + "inputs": [ + "transpose_62" + ], + "output": "slice_131", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_132", + "inputs": [ + "transpose_62" + ], + "output": "slice_132", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_133", + "inputs": [ + "transpose_62" + ], + "output": "slice_133", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_14", + "inputs": [ + "linear_111" + ], + "output": "sigmoid_14", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_11", + "inputs": [ + "p_model_model_language_model_layers_14_linear_attn_a_log" + ], + "output": "exp_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_17", + "inputs": [ + "exp_11" + ], + "output": "neg_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_151", + "inputs": [ + "linear_112", + "p_model_model_language_model_layers_14_linear_attn_dt_bias" + ], + "output": "add_151", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_11", + "inputs": [ + "add_151" + ], + "output": "softplus_11", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_192", + "inputs": [ + "neg_17", + "softplus_11" + ], + "output": "mul_192", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_193", + "inputs": [ + "slice_131" + ], + "output": "mul_193", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_23", + "inputs": [ + "mul_193" + ], + "output": "sum_23", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_152", + "inputs": [ + "sum_23" + ], + "output": "add_152", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_68", + "inputs": [ + "add_152" + ], + "output": "rsqrt_68", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_194", + "inputs": [ + "slice_131", + "rsqrt_68" + ], + "output": "mul_194", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_195", + "inputs": [ + "slice_132" + ], + "output": "mul_195", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_24", + "inputs": [ + "mul_195" + ], + "output": "sum_24", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_153", + "inputs": [ + "sum_24" + ], + "output": "add_153", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_69", + "inputs": [ + "add_153" + ], + "output": "rsqrt_69", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_196", + "inputs": [ + "slice_132", + "rsqrt_69" + ], + "output": "mul_196", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_11", + "inputs": [ + "mul_194", + "mul_196", + "slice_133", + "mul_192", + "sigmoid_14", + "states_29" + ], + "output": "normalized_gated_delta_rule_11", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_134", + "inputs": [ + "normalized_gated_delta_rule_11" + ], + "output": "slice_134", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_29" + }, + { + "node": "slice_135", + "inputs": [ + "normalized_gated_delta_rule_11" + ], + "output": "slice_135", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_63", + "inputs": [ + "slice_135" + ], + "output": "transpose_63", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_198", + "inputs": [ + "transpose_63", + "p_model_model_language_model_layers_14_linear_attn_norm_weight" + ], + "output": "mul_198", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_199", + "inputs": [ + "linear_110", + "mul_198" + ], + "output": "mul_199", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_113", + "inputs": [ + "mul_199", + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_113", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_155", + "inputs": [ + "add_148", + "linear_113" + ], + "output": "add_155", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_201", + "inputs": [ + "add_155", + "p_model_model_language_model_layers_14_post_attention_layernorm_weight" + ], + "output": "mul_201", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_114", + "inputs": [ + "mul_201", + "p_model_model_language_model_layers_14_mlp_gate_proj_weight", + "p_model_model_language_model_layers_14_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_114", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_115", + "inputs": [ + "mul_201", + "p_model_model_language_model_layers_14_mlp_up_proj_weight", + "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_115", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_202", + "inputs": [ + "linear_114", + "linear_115" + ], + "output": "mul_202", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_116", + "inputs": [ + "mul_202", + "p_model_model_language_model_layers_14_mlp_down_proj_weight", + "p_model_model_language_model_layers_14_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_116", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_158", + "inputs": [ + "add_155", + "linear_116" + ], + "output": "add_158", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_204", + "inputs": [ + "add_158", + "p_model_model_language_model_layers_15_input_layernorm_weight" + ], + "output": "mul_204", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_117", + "inputs": [ + "mul_204", + "p_model_model_language_model_layers_15_self_attn_q_proj_weight", + "p_model_model_language_model_layers_15_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_117", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/8b540ef0dbe320e8ccc513d0b118e62c6d25dff607a62b8d9de03d1a393a0a5c.wgsl" + }, + { + "node": "slice_136", + "inputs": [ + "linear_117" + ], + "output": "slice_136", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/9d3d7b777701cd320ca11358e554e1019fb041ee4d765e66071c215242906773.wgsl" + }, + { + "node": "slice_137", + "inputs": [ + "linear_117" + ], + "output": "slice_137", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/3622ffa2072335a35d6c9db5d284ed92b3fb5ed304acc2c606408063e7e21dcd.wgsl" + }, + { + "node": "mul_206", + "inputs": [ + "slice_136", + "p_model_model_language_model_layers_15_self_attn_q_norm_weight" + ], + "output": "mul_206", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl" + }, + { + "node": "transpose_64", + "inputs": [ + "mul_206" + ], + "output": "transpose_64", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/63396da046b345383e1c7b09aee551d08e4c3c45030ff5c8bb76bbc34bb2abc6.wgsl" + }, + { + "node": "linear_118", + "inputs": [ + "mul_204", + "p_model_model_language_model_layers_15_self_attn_k_proj_weight", + "p_model_model_language_model_layers_15_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_118", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl" + }, + { + "node": "mul_208", + "inputs": [ + "linear_118", + "p_model_model_language_model_layers_15_self_attn_k_norm_weight" + ], + "output": "mul_208", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl" + }, + { + "node": "transpose_65", + "inputs": [ + "mul_208" + ], + "output": "transpose_65", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl" + }, + { + "node": "linear_119", + "inputs": [ + "mul_204", + "p_model_model_language_model_layers_15_self_attn_v_proj_weight", + "p_model_model_language_model_layers_15_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_119", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl" + }, + { + "node": "transpose_66", + "inputs": [ + "linear_119" + ], + "output": "transpose_66", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl" + }, + { + "node": "slice_138", + "inputs": [ + "transpose_64" + ], + "output": "slice_138", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f2ee90b3d949141c23610237d27d44a7b833aacae8cdbc2e42a595d5eb7b5537.wgsl" + }, + { + "node": "slice_139", + "inputs": [ + "transpose_64" + ], + "output": "slice_139", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/8d62df48089497dd19fea48680730210decda42b63f69ecdf9688fc4a43efc58.wgsl" + }, + { + "node": "slice_140", + "inputs": [ + "transpose_65" + ], + "output": "slice_140", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6caa35fcab56c639d1859e1b090a7e5bf016dd7f33ecd2373c24ab625de16e09.wgsl" + }, + { + "node": "slice_141", + "inputs": [ + "transpose_65" + ], + "output": "slice_141", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/2182574f59b7d9e38d74b30540c5781827e7da2de1ba6b00327af0b4cb42379d.wgsl" + }, + { + "node": "mul_209", + "inputs": [ + "slice_138", + "mul" + ], + "output": "mul_209", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e3c6a64e01b6ea79adf125ff00a6dcf00f0b65aee0f4a3d14fdbc20e5dfeb22d.wgsl" + }, + { + "node": "slice_142", + "inputs": [ + "slice_138" + ], + "output": "slice_142", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/b5984ff6eb9e56ef0f9930f775c1dc3c203eec15924e5fea4c2525b2ce180857.wgsl" + }, + { + "node": "slice_143", + "inputs": [ + "slice_138" + ], + "output": "slice_143", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/8256ea6cdc3500306470c0bb2dcdb44ebdd454dafa218ad6a94dd953b9bbe125.wgsl" + }, + { + "node": "neg_18", + "inputs": [ + "slice_143" + ], + "output": "neg_18", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/adf1e58a124e1301625024ae142dca5fb1dd50799fa94b52eaeffb82ea99852c.wgsl" + }, + { + "node": "cat_25", + "inputs": [ + "neg_18", + "slice_142" + ], + "output": "cat_25", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl" + }, + { + "node": "mul_210", + "inputs": [ + "cat_25", + "mul_1" + ], + "output": "mul_210", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e3c6a64e01b6ea79adf125ff00a6dcf00f0b65aee0f4a3d14fdbc20e5dfeb22d.wgsl" + }, + { + "node": "add_165", + "inputs": [ + "mul_209", + "mul_210" + ], + "output": "add_165", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl" + }, + { + "node": "mul_211", + "inputs": [ + "slice_140", + "mul" + ], + "output": "mul_211", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl" + }, + { + "node": "slice_144", + "inputs": [ + "slice_140" + ], + "output": "slice_144", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/58077eb80a5fa253f535c97c82dc07edd48455a17cc52583aeee41c7d1bbd0b3.wgsl" + }, + { + "node": "slice_145", + "inputs": [ + "slice_140" + ], + "output": "slice_145", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3830061b0fc54c85d3d8a783b927c44484b1a3b68b52b8d8a1b980f7553deebf.wgsl" + }, + { + "node": "neg_19", + "inputs": [ + "slice_145" + ], + "output": "neg_19", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl" + }, + { + "node": "cat_26", + "inputs": [ + "neg_19", + "slice_144" + ], + "output": "cat_26", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/bda7ebd3bcf25c31938fca1a582e82b972ab742f85f826f085417b0e6f651e60.wgsl" + }, + { + "node": "mul_212", + "inputs": [ + "cat_26", + "mul_1" + ], + "output": "mul_212", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl" + }, + { + "node": "add_166", + "inputs": [ + "mul_211", + "mul_212" + ], + "output": "add_166", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl" + }, + { + "node": "cat_27", + "inputs": [ + "add_165", + "slice_139" + ], + "output": "cat_27", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl" + }, + { + "node": "cat_28", + "inputs": [ + "add_166", + "slice_141" + ], + "output": "cat_28", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl" + }, + { + "node": "cache_update_6", + "inputs": [ + "cat_28", + "position_ids" + ], + "output": "cache_update_6", + "workgroups": [ + 128, + 1, + 1 + ], + "inPlaceOf": "states_30", + "kernel": "kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl" + }, + { + "node": "cache_update_7", + "inputs": [ + "transpose_66", + "position_ids" + ], + "output": "cache_update_7", + "workgroups": [ + 128, + 1, + 1 + ], + "inPlaceOf": "states_31", + "kernel": "kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl" + }, + { + "node": "matmul_8", + "inputs": [ + "cat_27", + "cache_update_6", + "cache_update_7", + "attention_mask" + ], + "output": "matmul_8", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d9ab25f0ef4248028d7dab1e782d0756dfc664cc90794d13644bd91106541608.wgsl" + }, + { + "node": "transpose_68", + "inputs": [ + "matmul_8" + ], + "output": "transpose_68", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/84f24861effe9e2f3b05f98fe6d4e16f4f8204be40a7eeece0668e57060ecc7b.wgsl" + }, + { + "node": "sigmoid_15", + "inputs": [ + "slice_137" + ], + "output": "sigmoid_15", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/816861e23c92e156c1ac01f21e4f6dd1cbfc14978b1be8a5994fb45b0f845cb8.wgsl" + }, + { + "node": "mul_214", + "inputs": [ + "transpose_68", + "sigmoid_15" + ], + "output": "mul_214", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/aa227ae49fab3f75d86e02e6f390be785eabfeafe70e77a5c7eb4ad51895e382.wgsl" + }, + { + "node": "linear_120", + "inputs": [ + "mul_214", + "p_model_model_language_model_layers_15_self_attn_o_proj_weight", + "p_model_model_language_model_layers_15_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_120", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_168", + "inputs": [ + "add_158", + "linear_120" + ], + "output": "add_168", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_216", + "inputs": [ + "add_168", + "p_model_model_language_model_layers_15_post_attention_layernorm_weight" + ], + "output": "mul_216", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_121", + "inputs": [ + "mul_216", + "p_model_model_language_model_layers_15_mlp_gate_proj_weight", + "p_model_model_language_model_layers_15_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_121", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_122", + "inputs": [ + "mul_216", + "p_model_model_language_model_layers_15_mlp_up_proj_weight", + "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_122", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_217", + "inputs": [ + "linear_121", + "linear_122" + ], + "output": "mul_217", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_123", + "inputs": [ + "mul_217", + "p_model_model_language_model_layers_15_mlp_down_proj_weight", + "p_model_model_language_model_layers_15_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_123", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_171", + "inputs": [ + "add_168", + "linear_123" + ], + "output": "add_171", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_219", + "inputs": [ + "add_171", + "p_model_model_language_model_layers_16_input_layernorm_weight" + ], + "output": "mul_219", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_124", + "inputs": [ + "mul_219", + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_124", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_69", + "inputs": [ + "linear_124" + ], + "output": "transpose_69", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_125", + "inputs": [ + "mul_219", + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_125", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_126", + "inputs": [ + "mul_219", + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_126", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_127", + "inputs": [ + "mul_219", + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_127", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_29", + "inputs": [ + "states_32", + "transpose_69" + ], + "output": "cat_29", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_146", + "inputs": [ + "cat_29" + ], + "output": "slice_146", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_32" + }, + { + "node": "conv1d_12", + "inputs": [ + "cat_29", + "p_model_model_language_model_layers_16_linear_attn_conv1d_weight" + ], + "output": "conv1d_12", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_147", + "inputs": [ + "conv1d_12" + ], + "output": "slice_147", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_40", + "inputs": [ + "slice_147" + ], + "output": "silu_40", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_148", + "inputs": [ + "silu_40" + ], + "output": "slice_148", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_70", + "inputs": [ + "slice_148" + ], + "output": "transpose_70", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_149", + "inputs": [ + "transpose_70" + ], + "output": "slice_149", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_150", + "inputs": [ + "transpose_70" + ], + "output": "slice_150", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_151", + "inputs": [ + "transpose_70" + ], + "output": "slice_151", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_16", + "inputs": [ + "linear_126" + ], + "output": "sigmoid_16", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_12", + "inputs": [ + "p_model_model_language_model_layers_16_linear_attn_a_log" + ], + "output": "exp_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_20", + "inputs": [ + "exp_12" + ], + "output": "neg_20", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_174", + "inputs": [ + "linear_127", + "p_model_model_language_model_layers_16_linear_attn_dt_bias" + ], + "output": "add_174", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_12", + "inputs": [ + "add_174" + ], + "output": "softplus_12", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_220", + "inputs": [ + "neg_20", + "softplus_12" + ], + "output": "mul_220", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_221", + "inputs": [ + "slice_149" + ], + "output": "mul_221", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_25", + "inputs": [ + "mul_221" + ], + "output": "sum_25", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_175", + "inputs": [ + "sum_25" + ], + "output": "add_175", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_77", + "inputs": [ + "add_175" + ], + "output": "rsqrt_77", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_222", + "inputs": [ + "slice_149", + "rsqrt_77" + ], + "output": "mul_222", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_223", + "inputs": [ + "slice_150" + ], + "output": "mul_223", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_26", + "inputs": [ + "mul_223" + ], + "output": "sum_26", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_176", + "inputs": [ + "sum_26" + ], + "output": "add_176", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_78", + "inputs": [ + "add_176" + ], + "output": "rsqrt_78", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_224", + "inputs": [ + "slice_150", + "rsqrt_78" + ], + "output": "mul_224", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_12", + "inputs": [ + "mul_222", + "mul_224", + "slice_151", + "mul_220", + "sigmoid_16", + "states_33" + ], + "output": "normalized_gated_delta_rule_12", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_152", + "inputs": [ + "normalized_gated_delta_rule_12" + ], + "output": "slice_152", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_33" + }, + { + "node": "slice_153", + "inputs": [ + "normalized_gated_delta_rule_12" + ], + "output": "slice_153", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_71", + "inputs": [ + "slice_153" + ], + "output": "transpose_71", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_226", + "inputs": [ + "transpose_71", + "p_model_model_language_model_layers_16_linear_attn_norm_weight" + ], + "output": "mul_226", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_227", + "inputs": [ + "linear_125", + "mul_226" + ], + "output": "mul_227", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_128", + "inputs": [ + "mul_227", + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_128", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_178", + "inputs": [ + "add_171", + "linear_128" + ], + "output": "add_178", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_229", + "inputs": [ + "add_178", + "p_model_model_language_model_layers_16_post_attention_layernorm_weight" + ], + "output": "mul_229", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_129", + "inputs": [ + "mul_229", + "p_model_model_language_model_layers_16_mlp_gate_proj_weight", + "p_model_model_language_model_layers_16_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_129", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_130", + "inputs": [ + "mul_229", + "p_model_model_language_model_layers_16_mlp_up_proj_weight", + "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_130", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_230", + "inputs": [ + "linear_129", + "linear_130" + ], + "output": "mul_230", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_131", + "inputs": [ + "mul_230", + "p_model_model_language_model_layers_16_mlp_down_proj_weight", + "p_model_model_language_model_layers_16_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_131", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_181", + "inputs": [ + "add_178", + "linear_131" + ], + "output": "add_181", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_232", + "inputs": [ + "add_181", + "p_model_model_language_model_layers_17_input_layernorm_weight" + ], + "output": "mul_232", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_132", + "inputs": [ + "mul_232", + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_132", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_72", + "inputs": [ + "linear_132" + ], + "output": "transpose_72", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_133", + "inputs": [ + "mul_232", + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_133", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_134", + "inputs": [ + "mul_232", + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_134", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_135", + "inputs": [ + "mul_232", + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_135", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_30", + "inputs": [ + "states_34", + "transpose_72" + ], + "output": "cat_30", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_154", + "inputs": [ + "cat_30" + ], + "output": "slice_154", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_34" + }, + { + "node": "conv1d_13", + "inputs": [ + "cat_30", + "p_model_model_language_model_layers_17_linear_attn_conv1d_weight" + ], + "output": "conv1d_13", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_155", + "inputs": [ + "conv1d_13" + ], + "output": "slice_155", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_43", + "inputs": [ + "slice_155" + ], + "output": "silu_43", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_156", + "inputs": [ + "silu_43" + ], + "output": "slice_156", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_73", + "inputs": [ + "slice_156" + ], + "output": "transpose_73", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_157", + "inputs": [ + "transpose_73" + ], + "output": "slice_157", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_158", + "inputs": [ + "transpose_73" + ], + "output": "slice_158", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_159", + "inputs": [ + "transpose_73" + ], + "output": "slice_159", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_17", + "inputs": [ + "linear_134" + ], + "output": "sigmoid_17", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_13", + "inputs": [ + "p_model_model_language_model_layers_17_linear_attn_a_log" + ], + "output": "exp_13", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_21", + "inputs": [ + "exp_13" + ], + "output": "neg_21", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_184", + "inputs": [ + "linear_135", + "p_model_model_language_model_layers_17_linear_attn_dt_bias" + ], + "output": "add_184", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_13", + "inputs": [ + "add_184" + ], + "output": "softplus_13", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_233", + "inputs": [ + "neg_21", + "softplus_13" + ], + "output": "mul_233", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_234", + "inputs": [ + "slice_157" + ], + "output": "mul_234", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_27", + "inputs": [ + "mul_234" + ], + "output": "sum_27", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_185", + "inputs": [ + "sum_27" + ], + "output": "add_185", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_82", + "inputs": [ + "add_185" + ], + "output": "rsqrt_82", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_235", + "inputs": [ + "slice_157", + "rsqrt_82" + ], + "output": "mul_235", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_236", + "inputs": [ + "slice_158" + ], + "output": "mul_236", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_28", + "inputs": [ + "mul_236" + ], + "output": "sum_28", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_186", + "inputs": [ + "sum_28" + ], + "output": "add_186", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_83", + "inputs": [ + "add_186" + ], + "output": "rsqrt_83", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_237", + "inputs": [ + "slice_158", + "rsqrt_83" + ], + "output": "mul_237", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_13", + "inputs": [ + "mul_235", + "mul_237", + "slice_159", + "mul_233", + "sigmoid_17", + "states_35" + ], + "output": "normalized_gated_delta_rule_13", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_160", + "inputs": [ + "normalized_gated_delta_rule_13" + ], + "output": "slice_160", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_35" + }, + { + "node": "slice_161", + "inputs": [ + "normalized_gated_delta_rule_13" + ], + "output": "slice_161", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_74", + "inputs": [ + "slice_161" + ], + "output": "transpose_74", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_239", + "inputs": [ + "transpose_74", + "p_model_model_language_model_layers_17_linear_attn_norm_weight" + ], + "output": "mul_239", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_240", + "inputs": [ + "linear_133", + "mul_239" + ], + "output": "mul_240", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_136", + "inputs": [ + "mul_240", + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_136", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_188", + "inputs": [ + "add_181", + "linear_136" + ], + "output": "add_188", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_242", + "inputs": [ + "add_188", + "p_model_model_language_model_layers_17_post_attention_layernorm_weight" + ], + "output": "mul_242", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_137", + "inputs": [ + "mul_242", + "p_model_model_language_model_layers_17_mlp_gate_proj_weight", + "p_model_model_language_model_layers_17_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_137", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_138", + "inputs": [ + "mul_242", + "p_model_model_language_model_layers_17_mlp_up_proj_weight", + "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_138", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_243", + "inputs": [ + "linear_137", + "linear_138" + ], + "output": "mul_243", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_139", + "inputs": [ + "mul_243", + "p_model_model_language_model_layers_17_mlp_down_proj_weight", + "p_model_model_language_model_layers_17_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_139", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_191", + "inputs": [ + "add_188", + "linear_139" + ], + "output": "add_191", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_245", + "inputs": [ + "add_191", + "p_model_model_language_model_layers_18_input_layernorm_weight" + ], + "output": "mul_245", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_140", + "inputs": [ + "mul_245", + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_140", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_75", + "inputs": [ + "linear_140" + ], + "output": "transpose_75", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_141", + "inputs": [ + "mul_245", + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_141", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_142", + "inputs": [ + "mul_245", + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_142", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_143", + "inputs": [ + "mul_245", + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_143", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_31", + "inputs": [ + "states_36", + "transpose_75" + ], + "output": "cat_31", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_162", + "inputs": [ + "cat_31" + ], + "output": "slice_162", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_36" + }, + { + "node": "conv1d_14", + "inputs": [ + "cat_31", + "p_model_model_language_model_layers_18_linear_attn_conv1d_weight" + ], + "output": "conv1d_14", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_163", + "inputs": [ + "conv1d_14" + ], + "output": "slice_163", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_46", + "inputs": [ + "slice_163" + ], + "output": "silu_46", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_164", + "inputs": [ + "silu_46" + ], + "output": "slice_164", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_76", + "inputs": [ + "slice_164" + ], + "output": "transpose_76", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_165", + "inputs": [ + "transpose_76" + ], + "output": "slice_165", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_166", + "inputs": [ + "transpose_76" + ], + "output": "slice_166", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_167", + "inputs": [ + "transpose_76" + ], + "output": "slice_167", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_18", + "inputs": [ + "linear_142" + ], + "output": "sigmoid_18", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_14", + "inputs": [ + "p_model_model_language_model_layers_18_linear_attn_a_log" + ], + "output": "exp_14", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_22", + "inputs": [ + "exp_14" + ], + "output": "neg_22", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_194", + "inputs": [ + "linear_143", + "p_model_model_language_model_layers_18_linear_attn_dt_bias" + ], + "output": "add_194", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_14", + "inputs": [ + "add_194" + ], + "output": "softplus_14", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_246", + "inputs": [ + "neg_22", + "softplus_14" + ], + "output": "mul_246", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_247", + "inputs": [ + "slice_165" + ], + "output": "mul_247", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_29", + "inputs": [ + "mul_247" + ], + "output": "sum_29", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_195", + "inputs": [ + "sum_29" + ], + "output": "add_195", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_87", + "inputs": [ + "add_195" + ], + "output": "rsqrt_87", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_248", + "inputs": [ + "slice_165", + "rsqrt_87" + ], + "output": "mul_248", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_249", + "inputs": [ + "slice_166" + ], + "output": "mul_249", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_30", + "inputs": [ + "mul_249" + ], + "output": "sum_30", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_196", + "inputs": [ + "sum_30" + ], + "output": "add_196", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_88", + "inputs": [ + "add_196" + ], + "output": "rsqrt_88", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_250", + "inputs": [ + "slice_166", + "rsqrt_88" + ], + "output": "mul_250", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_14", + "inputs": [ + "mul_248", + "mul_250", + "slice_167", + "mul_246", + "sigmoid_18", + "states_37" + ], + "output": "normalized_gated_delta_rule_14", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_168", + "inputs": [ + "normalized_gated_delta_rule_14" + ], + "output": "slice_168", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_37" + }, + { + "node": "slice_169", + "inputs": [ + "normalized_gated_delta_rule_14" + ], + "output": "slice_169", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_77", + "inputs": [ + "slice_169" + ], + "output": "transpose_77", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_252", + "inputs": [ + "transpose_77", + "p_model_model_language_model_layers_18_linear_attn_norm_weight" + ], + "output": "mul_252", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_253", + "inputs": [ + "linear_141", + "mul_252" + ], + "output": "mul_253", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_144", + "inputs": [ + "mul_253", + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_144", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_198", + "inputs": [ + "add_191", + "linear_144" + ], + "output": "add_198", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_255", + "inputs": [ + "add_198", + "p_model_model_language_model_layers_18_post_attention_layernorm_weight" + ], + "output": "mul_255", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_145", + "inputs": [ + "mul_255", + "p_model_model_language_model_layers_18_mlp_gate_proj_weight", + "p_model_model_language_model_layers_18_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_145", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_146", + "inputs": [ + "mul_255", + "p_model_model_language_model_layers_18_mlp_up_proj_weight", + "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_146", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_256", + "inputs": [ + "linear_145", + "linear_146" + ], + "output": "mul_256", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_147", + "inputs": [ + "mul_256", + "p_model_model_language_model_layers_18_mlp_down_proj_weight", + "p_model_model_language_model_layers_18_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_147", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_201", + "inputs": [ + "add_198", + "linear_147" + ], + "output": "add_201", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_258", + "inputs": [ + "add_201", + "p_model_model_language_model_layers_19_input_layernorm_weight" + ], + "output": "mul_258", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_148", + "inputs": [ + "mul_258", + "p_model_model_language_model_layers_19_self_attn_q_proj_weight", + "p_model_model_language_model_layers_19_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_148", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/8b540ef0dbe320e8ccc513d0b118e62c6d25dff607a62b8d9de03d1a393a0a5c.wgsl" + }, + { + "node": "slice_170", + "inputs": [ + "linear_148" + ], + "output": "slice_170", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/9d3d7b777701cd320ca11358e554e1019fb041ee4d765e66071c215242906773.wgsl" + }, + { + "node": "slice_171", + "inputs": [ + "linear_148" + ], + "output": "slice_171", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/3622ffa2072335a35d6c9db5d284ed92b3fb5ed304acc2c606408063e7e21dcd.wgsl" + }, + { + "node": "mul_260", + "inputs": [ + "slice_170", + "p_model_model_language_model_layers_19_self_attn_q_norm_weight" + ], + "output": "mul_260", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl" + }, + { + "node": "transpose_78", + "inputs": [ + "mul_260" + ], + "output": "transpose_78", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/63396da046b345383e1c7b09aee551d08e4c3c45030ff5c8bb76bbc34bb2abc6.wgsl" + }, + { + "node": "linear_149", + "inputs": [ + "mul_258", + "p_model_model_language_model_layers_19_self_attn_k_proj_weight", + "p_model_model_language_model_layers_19_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_149", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl" + }, + { + "node": "mul_262", + "inputs": [ + "linear_149", + "p_model_model_language_model_layers_19_self_attn_k_norm_weight" + ], + "output": "mul_262", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl" + }, + { + "node": "transpose_79", + "inputs": [ + "mul_262" + ], + "output": "transpose_79", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl" + }, + { + "node": "linear_150", + "inputs": [ + "mul_258", + "p_model_model_language_model_layers_19_self_attn_v_proj_weight", + "p_model_model_language_model_layers_19_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_150", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl" + }, + { + "node": "transpose_80", + "inputs": [ + "linear_150" + ], + "output": "transpose_80", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl" + }, + { + "node": "slice_172", + "inputs": [ + "transpose_78" + ], + "output": "slice_172", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f2ee90b3d949141c23610237d27d44a7b833aacae8cdbc2e42a595d5eb7b5537.wgsl" + }, + { + "node": "slice_173", + "inputs": [ + "transpose_78" + ], + "output": "slice_173", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/8d62df48089497dd19fea48680730210decda42b63f69ecdf9688fc4a43efc58.wgsl" + }, + { + "node": "slice_174", + "inputs": [ + "transpose_79" + ], + "output": "slice_174", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6caa35fcab56c639d1859e1b090a7e5bf016dd7f33ecd2373c24ab625de16e09.wgsl" + }, + { + "node": "slice_175", + "inputs": [ + "transpose_79" + ], + "output": "slice_175", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/2182574f59b7d9e38d74b30540c5781827e7da2de1ba6b00327af0b4cb42379d.wgsl" + }, + { + "node": "mul_263", + "inputs": [ + "slice_172", + "mul" + ], + "output": "mul_263", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e3c6a64e01b6ea79adf125ff00a6dcf00f0b65aee0f4a3d14fdbc20e5dfeb22d.wgsl" + }, + { + "node": "slice_176", + "inputs": [ + "slice_172" + ], + "output": "slice_176", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/b5984ff6eb9e56ef0f9930f775c1dc3c203eec15924e5fea4c2525b2ce180857.wgsl" + }, + { + "node": "slice_177", + "inputs": [ + "slice_172" + ], + "output": "slice_177", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/8256ea6cdc3500306470c0bb2dcdb44ebdd454dafa218ad6a94dd953b9bbe125.wgsl" + }, + { + "node": "neg_23", + "inputs": [ + "slice_177" + ], + "output": "neg_23", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/adf1e58a124e1301625024ae142dca5fb1dd50799fa94b52eaeffb82ea99852c.wgsl" + }, + { + "node": "cat_32", + "inputs": [ + "neg_23", + "slice_176" + ], + "output": "cat_32", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl" + }, + { + "node": "mul_264", + "inputs": [ + "cat_32", + "mul_1" + ], + "output": "mul_264", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e3c6a64e01b6ea79adf125ff00a6dcf00f0b65aee0f4a3d14fdbc20e5dfeb22d.wgsl" + }, + { + "node": "add_208", + "inputs": [ + "mul_263", + "mul_264" + ], + "output": "add_208", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl" + }, + { + "node": "mul_265", + "inputs": [ + "slice_174", + "mul" + ], + "output": "mul_265", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl" + }, + { + "node": "slice_178", + "inputs": [ + "slice_174" + ], + "output": "slice_178", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/58077eb80a5fa253f535c97c82dc07edd48455a17cc52583aeee41c7d1bbd0b3.wgsl" + }, + { + "node": "slice_179", + "inputs": [ + "slice_174" + ], + "output": "slice_179", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3830061b0fc54c85d3d8a783b927c44484b1a3b68b52b8d8a1b980f7553deebf.wgsl" + }, + { + "node": "neg_24", + "inputs": [ + "slice_179" + ], + "output": "neg_24", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl" + }, + { + "node": "cat_33", + "inputs": [ + "neg_24", + "slice_178" + ], + "output": "cat_33", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/bda7ebd3bcf25c31938fca1a582e82b972ab742f85f826f085417b0e6f651e60.wgsl" + }, + { + "node": "mul_266", + "inputs": [ + "cat_33", + "mul_1" + ], + "output": "mul_266", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl" + }, + { + "node": "add_209", + "inputs": [ + "mul_265", + "mul_266" + ], + "output": "add_209", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl" + }, + { + "node": "cat_34", + "inputs": [ + "add_208", + "slice_173" + ], + "output": "cat_34", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl" + }, + { + "node": "cat_35", + "inputs": [ + "add_209", + "slice_175" + ], + "output": "cat_35", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl" + }, + { + "node": "cache_update_8", + "inputs": [ + "cat_35", + "position_ids" + ], + "output": "cache_update_8", + "workgroups": [ + 128, + 1, + 1 + ], + "inPlaceOf": "states_38", + "kernel": "kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl" + }, + { + "node": "cache_update_9", + "inputs": [ + "transpose_80", + "position_ids" + ], + "output": "cache_update_9", + "workgroups": [ + 128, + 1, + 1 + ], + "inPlaceOf": "states_39", + "kernel": "kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl" + }, + { + "node": "matmul_10", + "inputs": [ + "cat_34", + "cache_update_8", + "cache_update_9", + "attention_mask" + ], + "output": "matmul_10", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d9ab25f0ef4248028d7dab1e782d0756dfc664cc90794d13644bd91106541608.wgsl" + }, + { + "node": "transpose_82", + "inputs": [ + "matmul_10" + ], + "output": "transpose_82", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/84f24861effe9e2f3b05f98fe6d4e16f4f8204be40a7eeece0668e57060ecc7b.wgsl" + }, + { + "node": "sigmoid_19", + "inputs": [ + "slice_171" + ], + "output": "sigmoid_19", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/816861e23c92e156c1ac01f21e4f6dd1cbfc14978b1be8a5994fb45b0f845cb8.wgsl" + }, + { + "node": "mul_268", + "inputs": [ + "transpose_82", + "sigmoid_19" + ], + "output": "mul_268", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/aa227ae49fab3f75d86e02e6f390be785eabfeafe70e77a5c7eb4ad51895e382.wgsl" + }, + { + "node": "linear_151", + "inputs": [ + "mul_268", + "p_model_model_language_model_layers_19_self_attn_o_proj_weight", + "p_model_model_language_model_layers_19_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_151", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_211", + "inputs": [ + "add_201", + "linear_151" + ], + "output": "add_211", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_270", + "inputs": [ + "add_211", + "p_model_model_language_model_layers_19_post_attention_layernorm_weight" + ], + "output": "mul_270", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_152", + "inputs": [ + "mul_270", + "p_model_model_language_model_layers_19_mlp_gate_proj_weight", + "p_model_model_language_model_layers_19_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_152", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_153", + "inputs": [ + "mul_270", + "p_model_model_language_model_layers_19_mlp_up_proj_weight", + "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_153", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_271", + "inputs": [ + "linear_152", + "linear_153" + ], + "output": "mul_271", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_154", + "inputs": [ + "mul_271", + "p_model_model_language_model_layers_19_mlp_down_proj_weight", + "p_model_model_language_model_layers_19_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_154", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_214", + "inputs": [ + "add_211", + "linear_154" + ], + "output": "add_214", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_273", + "inputs": [ + "add_214", + "p_model_model_language_model_layers_20_input_layernorm_weight" + ], + "output": "mul_273", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_155", + "inputs": [ + "mul_273", + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_155", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_83", + "inputs": [ + "linear_155" + ], + "output": "transpose_83", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_156", + "inputs": [ + "mul_273", + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_156", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_157", + "inputs": [ + "mul_273", + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_157", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_158", + "inputs": [ + "mul_273", + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_158", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_36", + "inputs": [ + "states_40", + "transpose_83" + ], + "output": "cat_36", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_180", + "inputs": [ + "cat_36" + ], + "output": "slice_180", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_40" + }, + { + "node": "conv1d_15", + "inputs": [ + "cat_36", + "p_model_model_language_model_layers_20_linear_attn_conv1d_weight" + ], + "output": "conv1d_15", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_181", + "inputs": [ + "conv1d_15" + ], + "output": "slice_181", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_50", + "inputs": [ + "slice_181" + ], + "output": "silu_50", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_182", + "inputs": [ + "silu_50" + ], + "output": "slice_182", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_84", + "inputs": [ + "slice_182" + ], + "output": "transpose_84", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_183", + "inputs": [ + "transpose_84" + ], + "output": "slice_183", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_184", + "inputs": [ + "transpose_84" + ], + "output": "slice_184", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_185", + "inputs": [ + "transpose_84" + ], + "output": "slice_185", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_20", + "inputs": [ + "linear_157" + ], + "output": "sigmoid_20", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_15", + "inputs": [ + "p_model_model_language_model_layers_20_linear_attn_a_log" + ], + "output": "exp_15", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_25", + "inputs": [ + "exp_15" + ], + "output": "neg_25", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_217", + "inputs": [ + "linear_158", + "p_model_model_language_model_layers_20_linear_attn_dt_bias" + ], + "output": "add_217", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_15", + "inputs": [ + "add_217" + ], + "output": "softplus_15", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_274", + "inputs": [ + "neg_25", + "softplus_15" + ], + "output": "mul_274", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_275", + "inputs": [ + "slice_183" + ], + "output": "mul_275", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_31", + "inputs": [ + "mul_275" + ], + "output": "sum_31", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_218", + "inputs": [ + "sum_31" + ], + "output": "add_218", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_96", + "inputs": [ + "add_218" + ], + "output": "rsqrt_96", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_276", + "inputs": [ + "slice_183", + "rsqrt_96" + ], + "output": "mul_276", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_277", + "inputs": [ + "slice_184" + ], + "output": "mul_277", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_32", + "inputs": [ + "mul_277" + ], + "output": "sum_32", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_219", + "inputs": [ + "sum_32" + ], + "output": "add_219", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_97", + "inputs": [ + "add_219" + ], + "output": "rsqrt_97", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_278", + "inputs": [ + "slice_184", + "rsqrt_97" + ], + "output": "mul_278", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_15", + "inputs": [ + "mul_276", + "mul_278", + "slice_185", + "mul_274", + "sigmoid_20", + "states_41" + ], + "output": "normalized_gated_delta_rule_15", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_186", + "inputs": [ + "normalized_gated_delta_rule_15" + ], + "output": "slice_186", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_41" + }, + { + "node": "slice_187", + "inputs": [ + "normalized_gated_delta_rule_15" + ], + "output": "slice_187", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_85", + "inputs": [ + "slice_187" + ], + "output": "transpose_85", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_280", + "inputs": [ + "transpose_85", + "p_model_model_language_model_layers_20_linear_attn_norm_weight" + ], + "output": "mul_280", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_281", + "inputs": [ + "linear_156", + "mul_280" + ], + "output": "mul_281", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_159", + "inputs": [ + "mul_281", + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_159", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_221", + "inputs": [ + "add_214", + "linear_159" + ], + "output": "add_221", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_283", + "inputs": [ + "add_221", + "p_model_model_language_model_layers_20_post_attention_layernorm_weight" + ], + "output": "mul_283", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_160", + "inputs": [ + "mul_283", + "p_model_model_language_model_layers_20_mlp_gate_proj_weight", + "p_model_model_language_model_layers_20_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_160", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_161", + "inputs": [ + "mul_283", + "p_model_model_language_model_layers_20_mlp_up_proj_weight", + "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_161", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_284", + "inputs": [ + "linear_160", + "linear_161" + ], + "output": "mul_284", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_162", + "inputs": [ + "mul_284", + "p_model_model_language_model_layers_20_mlp_down_proj_weight", + "p_model_model_language_model_layers_20_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_162", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_224", + "inputs": [ + "add_221", + "linear_162" + ], + "output": "add_224", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_286", + "inputs": [ + "add_224", + "p_model_model_language_model_layers_21_input_layernorm_weight" + ], + "output": "mul_286", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_163", + "inputs": [ + "mul_286", + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_163", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_86", + "inputs": [ + "linear_163" + ], + "output": "transpose_86", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_164", + "inputs": [ + "mul_286", + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_164", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_165", + "inputs": [ + "mul_286", + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_165", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_166", + "inputs": [ + "mul_286", + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_166", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_37", + "inputs": [ + "states_42", + "transpose_86" + ], + "output": "cat_37", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_188", + "inputs": [ + "cat_37" + ], + "output": "slice_188", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_42" + }, + { + "node": "conv1d_16", + "inputs": [ + "cat_37", + "p_model_model_language_model_layers_21_linear_attn_conv1d_weight" + ], + "output": "conv1d_16", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_189", + "inputs": [ + "conv1d_16" + ], + "output": "slice_189", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_53", + "inputs": [ + "slice_189" + ], + "output": "silu_53", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_190", + "inputs": [ + "silu_53" + ], + "output": "slice_190", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_87", + "inputs": [ + "slice_190" + ], + "output": "transpose_87", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_191", + "inputs": [ + "transpose_87" + ], + "output": "slice_191", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_192", + "inputs": [ + "transpose_87" + ], + "output": "slice_192", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_193", + "inputs": [ + "transpose_87" + ], + "output": "slice_193", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_21", + "inputs": [ + "linear_165" + ], + "output": "sigmoid_21", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_16", + "inputs": [ + "p_model_model_language_model_layers_21_linear_attn_a_log" + ], + "output": "exp_16", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_26", + "inputs": [ + "exp_16" + ], + "output": "neg_26", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_227", + "inputs": [ + "linear_166", + "p_model_model_language_model_layers_21_linear_attn_dt_bias" + ], + "output": "add_227", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_16", + "inputs": [ + "add_227" + ], + "output": "softplus_16", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_287", + "inputs": [ + "neg_26", + "softplus_16" + ], + "output": "mul_287", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_288", + "inputs": [ + "slice_191" + ], + "output": "mul_288", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_33", + "inputs": [ + "mul_288" + ], + "output": "sum_33", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_228", + "inputs": [ + "sum_33" + ], + "output": "add_228", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_101", + "inputs": [ + "add_228" + ], + "output": "rsqrt_101", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_289", + "inputs": [ + "slice_191", + "rsqrt_101" + ], + "output": "mul_289", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_290", + "inputs": [ + "slice_192" + ], + "output": "mul_290", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_34", + "inputs": [ + "mul_290" + ], + "output": "sum_34", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_229", + "inputs": [ + "sum_34" + ], + "output": "add_229", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_102", + "inputs": [ + "add_229" + ], + "output": "rsqrt_102", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_291", + "inputs": [ + "slice_192", + "rsqrt_102" + ], + "output": "mul_291", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_16", + "inputs": [ + "mul_289", + "mul_291", + "slice_193", + "mul_287", + "sigmoid_21", + "states_43" + ], + "output": "normalized_gated_delta_rule_16", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_194", + "inputs": [ + "normalized_gated_delta_rule_16" + ], + "output": "slice_194", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_43" + }, + { + "node": "slice_195", + "inputs": [ + "normalized_gated_delta_rule_16" + ], + "output": "slice_195", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_88", + "inputs": [ + "slice_195" + ], + "output": "transpose_88", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_293", + "inputs": [ + "transpose_88", + "p_model_model_language_model_layers_21_linear_attn_norm_weight" + ], + "output": "mul_293", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_294", + "inputs": [ + "linear_164", + "mul_293" + ], + "output": "mul_294", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_167", + "inputs": [ + "mul_294", + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_167", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_231", + "inputs": [ + "add_224", + "linear_167" + ], + "output": "add_231", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_296", + "inputs": [ + "add_231", + "p_model_model_language_model_layers_21_post_attention_layernorm_weight" + ], + "output": "mul_296", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_168", + "inputs": [ + "mul_296", + "p_model_model_language_model_layers_21_mlp_gate_proj_weight", + "p_model_model_language_model_layers_21_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_168", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_169", + "inputs": [ + "mul_296", + "p_model_model_language_model_layers_21_mlp_up_proj_weight", + "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_169", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_297", + "inputs": [ + "linear_168", + "linear_169" + ], + "output": "mul_297", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_170", + "inputs": [ + "mul_297", + "p_model_model_language_model_layers_21_mlp_down_proj_weight", + "p_model_model_language_model_layers_21_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_170", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_234", + "inputs": [ + "add_231", + "linear_170" + ], + "output": "add_234", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_299", + "inputs": [ + "add_234", + "p_model_model_language_model_layers_22_input_layernorm_weight" + ], + "output": "mul_299", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_171", + "inputs": [ + "mul_299", + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_171", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/d0651e139bbc689d5d2ec68fb44d306820ef5d2a08443d6a2fb17774c3d8b6e1.wgsl" + }, + { + "node": "transpose_89", + "inputs": [ + "linear_171" + ], + "output": "transpose_89", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl" + }, + { + "node": "linear_172", + "inputs": [ + "mul_299", + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_172", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl" + }, + { + "node": "linear_173", + "inputs": [ + "mul_299", + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_173", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "linear_174", + "inputs": [ + "mul_299", + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_174", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/fcedaa0e15a5e700f580cd6605b1796a47d852dc3f3ef4ebf713f7b7292ae791.wgsl" + }, + { + "node": "cat_38", + "inputs": [ + "states_44", + "transpose_89" + ], + "output": "cat_38", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl" + }, + { + "node": "slice_196", + "inputs": [ + "cat_38" + ], + "output": "slice_196", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl", + "inPlaceOf": "states_44" + }, + { + "node": "conv1d_17", + "inputs": [ + "cat_38", + "p_model_model_language_model_layers_22_linear_attn_conv1d_weight" + ], + "output": "conv1d_17", + "workgroups": [ + 2208, + 1, + 1 + ], + "kernel": "kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl" + }, + { + "node": "slice_197", + "inputs": [ + "conv1d_17" + ], + "output": "slice_197", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl" + }, + { + "node": "silu_56", + "inputs": [ + "slice_197" + ], + "output": "silu_56", + "workgroups": [ + 1920, + 1, + 1 + ], + "kernel": "kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl" + }, + { + "node": "slice_198", + "inputs": [ + "silu_56" + ], + "output": "slice_198", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl" + }, + { + "node": "transpose_90", + "inputs": [ + "slice_198" + ], + "output": "transpose_90", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/a7b50e8b6e4d56404b6261383f50b3eed4a2a59c3fadaaffcc974575c32a374f.wgsl" + }, + { + "node": "slice_199", + "inputs": [ + "transpose_90" + ], + "output": "slice_199", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/c24c7e6057a3ab4778def9751ec5123025ec0e36139db626eccc4a9151bc6cc3.wgsl" + }, + { + "node": "slice_200", + "inputs": [ + "transpose_90" + ], + "output": "slice_200", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/f8460b61ea8ab5946764c8a8fa3a6aa4a51c4db1e30f5f1d3e3f0f2f1635a964.wgsl" + }, + { + "node": "slice_201", + "inputs": [ + "transpose_90" + ], + "output": "slice_201", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl" + }, + { + "node": "sigmoid_22", + "inputs": [ + "linear_173" + ], + "output": "sigmoid_22", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl" + }, + { + "node": "exp_17", + "inputs": [ + "p_model_model_language_model_layers_22_linear_attn_a_log" + ], + "output": "exp_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_27", + "inputs": [ + "exp_17" + ], + "output": "neg_27", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_237", + "inputs": [ + "linear_174", + "p_model_model_language_model_layers_22_linear_attn_dt_bias" + ], + "output": "add_237", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl" + }, + { + "node": "softplus_17", + "inputs": [ + "add_237" + ], + "output": "softplus_17", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl" + }, + { + "node": "mul_300", + "inputs": [ + "neg_27", + "softplus_17" + ], + "output": "mul_300", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/c02ac02d110c0be5d0990d20ab0d8c7a081c74f4e147f083789a203f22fc3258.wgsl" + }, + { + "node": "mul_301", + "inputs": [ + "slice_199" + ], + "output": "mul_301", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_35", + "inputs": [ + "mul_301" + ], + "output": "sum_35", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_238", + "inputs": [ + "sum_35" + ], + "output": "add_238", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_106", + "inputs": [ + "add_238" + ], + "output": "rsqrt_106", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_302", + "inputs": [ + "slice_199", + "rsqrt_106" + ], + "output": "mul_302", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "mul_303", + "inputs": [ + "slice_200" + ], + "output": "mul_303", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl" + }, + { + "node": "sum_36", + "inputs": [ + "mul_303" + ], + "output": "sum_36", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl" + }, + { + "node": "add_239", + "inputs": [ + "sum_36" + ], + "output": "add_239", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl" + }, + { + "node": "rsqrt_107", + "inputs": [ + "add_239" + ], + "output": "rsqrt_107", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl" + }, + { + "node": "mul_304", + "inputs": [ + "slice_200", + "rsqrt_107" + ], + "output": "mul_304", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl" + }, + { + "node": "normalized_gated_delta_rule_17", + "inputs": [ + "mul_302", + "mul_304", + "slice_201", + "mul_300", + "sigmoid_22", + "states_45" + ], + "output": "normalized_gated_delta_rule_17", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl" + }, + { + "node": "slice_202", + "inputs": [ + "normalized_gated_delta_rule_17" + ], + "output": "slice_202", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl", + "inPlaceOf": "states_45" + }, + { + "node": "slice_203", + "inputs": [ + "normalized_gated_delta_rule_17" + ], + "output": "slice_203", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl" + }, + { + "node": "transpose_91", + "inputs": [ + "slice_203" + ], + "output": "transpose_91", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl" + }, + { + "node": "mul_306", + "inputs": [ + "transpose_91", + "p_model_model_language_model_layers_22_linear_attn_norm_weight" + ], + "output": "mul_306", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl" + }, + { + "node": "mul_307", + "inputs": [ + "linear_172", + "mul_306" + ], + "output": "mul_307", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl" + }, + { + "node": "linear_175", + "inputs": [ + "mul_307", + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_175", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_241", + "inputs": [ + "add_234", + "linear_175" + ], + "output": "add_241", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_309", + "inputs": [ + "add_241", + "p_model_model_language_model_layers_22_post_attention_layernorm_weight" + ], + "output": "mul_309", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_176", + "inputs": [ + "mul_309", + "p_model_model_language_model_layers_22_mlp_gate_proj_weight", + "p_model_model_language_model_layers_22_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_176", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_177", + "inputs": [ + "mul_309", + "p_model_model_language_model_layers_22_mlp_up_proj_weight", + "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_177", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_310", + "inputs": [ + "linear_176", + "linear_177" + ], + "output": "mul_310", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_178", + "inputs": [ + "mul_310", + "p_model_model_language_model_layers_22_mlp_down_proj_weight", + "p_model_model_language_model_layers_22_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_178", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_244", + "inputs": [ + "add_241", + "linear_178" + ], + "output": "add_244", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_312", + "inputs": [ + "add_244", + "p_model_model_language_model_layers_23_input_layernorm_weight" + ], + "output": "mul_312", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_179", + "inputs": [ + "mul_312", + "p_model_model_language_model_layers_23_self_attn_q_proj_weight", + "p_model_model_language_model_layers_23_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_179", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/8b540ef0dbe320e8ccc513d0b118e62c6d25dff607a62b8d9de03d1a393a0a5c.wgsl" + }, + { + "node": "slice_204", + "inputs": [ + "linear_179" + ], + "output": "slice_204", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/9d3d7b777701cd320ca11358e554e1019fb041ee4d765e66071c215242906773.wgsl" + }, + { + "node": "slice_205", + "inputs": [ + "linear_179" + ], + "output": "slice_205", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/3622ffa2072335a35d6c9db5d284ed92b3fb5ed304acc2c606408063e7e21dcd.wgsl" + }, + { + "node": "mul_314", + "inputs": [ + "slice_204", + "p_model_model_language_model_layers_23_self_attn_q_norm_weight" + ], + "output": "mul_314", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl" + }, + { + "node": "transpose_92", + "inputs": [ + "mul_314" + ], + "output": "transpose_92", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/63396da046b345383e1c7b09aee551d08e4c3c45030ff5c8bb76bbc34bb2abc6.wgsl" + }, + { + "node": "linear_180", + "inputs": [ + "mul_312", + "p_model_model_language_model_layers_23_self_attn_k_proj_weight", + "p_model_model_language_model_layers_23_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_180", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl" + }, + { + "node": "mul_316", + "inputs": [ + "linear_180", + "p_model_model_language_model_layers_23_self_attn_k_norm_weight" + ], + "output": "mul_316", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl" + }, + { + "node": "transpose_93", + "inputs": [ + "mul_316" + ], + "output": "transpose_93", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl" + }, + { + "node": "linear_181", + "inputs": [ + "mul_312", + "p_model_model_language_model_layers_23_self_attn_v_proj_weight", + "p_model_model_language_model_layers_23_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_181", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl" + }, + { + "node": "transpose_94", + "inputs": [ + "linear_181" + ], + "output": "transpose_94", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl" + }, + { + "node": "slice_206", + "inputs": [ + "transpose_92" + ], + "output": "slice_206", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f2ee90b3d949141c23610237d27d44a7b833aacae8cdbc2e42a595d5eb7b5537.wgsl" + }, + { + "node": "slice_207", + "inputs": [ + "transpose_92" + ], + "output": "slice_207", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/8d62df48089497dd19fea48680730210decda42b63f69ecdf9688fc4a43efc58.wgsl" + }, + { + "node": "slice_208", + "inputs": [ + "transpose_93" + ], + "output": "slice_208", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/6caa35fcab56c639d1859e1b090a7e5bf016dd7f33ecd2373c24ab625de16e09.wgsl" + }, + { + "node": "slice_209", + "inputs": [ + "transpose_93" + ], + "output": "slice_209", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/2182574f59b7d9e38d74b30540c5781827e7da2de1ba6b00327af0b4cb42379d.wgsl" + }, + { + "node": "mul_317", + "inputs": [ + "slice_206", + "mul" + ], + "output": "mul_317", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e3c6a64e01b6ea79adf125ff00a6dcf00f0b65aee0f4a3d14fdbc20e5dfeb22d.wgsl" + }, + { + "node": "slice_210", + "inputs": [ + "slice_206" + ], + "output": "slice_210", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/b5984ff6eb9e56ef0f9930f775c1dc3c203eec15924e5fea4c2525b2ce180857.wgsl" + }, + { + "node": "slice_211", + "inputs": [ + "slice_206" + ], + "output": "slice_211", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/8256ea6cdc3500306470c0bb2dcdb44ebdd454dafa218ad6a94dd953b9bbe125.wgsl" + }, + { + "node": "neg_28", + "inputs": [ + "slice_211" + ], + "output": "neg_28", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/adf1e58a124e1301625024ae142dca5fb1dd50799fa94b52eaeffb82ea99852c.wgsl" + }, + { + "node": "cat_39", + "inputs": [ + "neg_28", + "slice_210" + ], + "output": "cat_39", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl" + }, + { + "node": "mul_318", + "inputs": [ + "cat_39", + "mul_1" + ], + "output": "mul_318", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e3c6a64e01b6ea79adf125ff00a6dcf00f0b65aee0f4a3d14fdbc20e5dfeb22d.wgsl" + }, + { + "node": "add_251", + "inputs": [ + "mul_317", + "mul_318" + ], + "output": "add_251", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl" + }, + { + "node": "mul_319", + "inputs": [ + "slice_208", + "mul" + ], + "output": "mul_319", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl" + }, + { + "node": "slice_212", + "inputs": [ + "slice_208" + ], + "output": "slice_212", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/58077eb80a5fa253f535c97c82dc07edd48455a17cc52583aeee41c7d1bbd0b3.wgsl" + }, + { + "node": "slice_213", + "inputs": [ + "slice_208" + ], + "output": "slice_213", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3830061b0fc54c85d3d8a783b927c44484b1a3b68b52b8d8a1b980f7553deebf.wgsl" + }, + { + "node": "neg_29", + "inputs": [ + "slice_213" + ], + "output": "neg_29", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl" + }, + { + "node": "cat_40", + "inputs": [ + "neg_29", + "slice_212" + ], + "output": "cat_40", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/bda7ebd3bcf25c31938fca1a582e82b972ab742f85f826f085417b0e6f651e60.wgsl" + }, + { + "node": "mul_320", + "inputs": [ + "cat_40", + "mul_1" + ], + "output": "mul_320", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl" + }, + { + "node": "add_252", + "inputs": [ + "mul_319", + "mul_320" + ], + "output": "add_252", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl" + }, + { + "node": "cat_41", + "inputs": [ + "add_251", + "slice_207" + ], + "output": "cat_41", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl" + }, + { + "node": "cat_42", + "inputs": [ + "add_252", + "slice_209" + ], + "output": "cat_42", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl" + }, + { + "node": "cache_update_10", + "inputs": [ + "cat_42", + "position_ids" + ], + "output": "cache_update_10", + "workgroups": [ + 128, + 1, + 1 + ], + "inPlaceOf": "states_46", + "kernel": "kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl" + }, + { + "node": "cache_update_11", + "inputs": [ + "transpose_94", + "position_ids" + ], + "output": "cache_update_11", + "workgroups": [ + 128, + 1, + 1 + ], + "inPlaceOf": "states_47", + "kernel": "kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl" + }, + { + "node": "matmul_12", + "inputs": [ + "cat_41", + "cache_update_10", + "cache_update_11", + "attention_mask" + ], + "output": "matmul_12", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/d9ab25f0ef4248028d7dab1e782d0756dfc664cc90794d13644bd91106541608.wgsl" + }, + { + "node": "transpose_96", + "inputs": [ + "matmul_12" + ], + "output": "transpose_96", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/84f24861effe9e2f3b05f98fe6d4e16f4f8204be40a7eeece0668e57060ecc7b.wgsl" + }, + { + "node": "sigmoid_23", + "inputs": [ + "slice_205" + ], + "output": "sigmoid_23", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/816861e23c92e156c1ac01f21e4f6dd1cbfc14978b1be8a5994fb45b0f845cb8.wgsl" + }, + { + "node": "mul_322", + "inputs": [ + "transpose_96", + "sigmoid_23" + ], + "output": "mul_322", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/aa227ae49fab3f75d86e02e6f390be785eabfeafe70e77a5c7eb4ad51895e382.wgsl" + }, + { + "node": "linear_182", + "inputs": [ + "mul_322", + "p_model_model_language_model_layers_23_self_attn_o_proj_weight", + "p_model_model_language_model_layers_23_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_182", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl" + }, + { + "node": "add_254", + "inputs": [ + "add_244", + "linear_182" + ], + "output": "add_254", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_324", + "inputs": [ + "add_254", + "p_model_model_language_model_layers_23_post_attention_layernorm_weight" + ], + "output": "mul_324", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_183", + "inputs": [ + "mul_324", + "p_model_model_language_model_layers_23_mlp_gate_proj_weight", + "p_model_model_language_model_layers_23_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_183", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "linear_184", + "inputs": [ + "mul_324", + "p_model_model_language_model_layers_23_mlp_up_proj_weight", + "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_184", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/d98c84cc12e2f8f4416252bb1d11dcb956262e8d672051da1e5c567f9671ab90.wgsl" + }, + { + "node": "mul_325", + "inputs": [ + "linear_183", + "linear_184" + ], + "output": "mul_325", + "workgroups": [ + 896, + 1, + 1 + ], + "kernel": "kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl" + }, + { + "node": "linear_185", + "inputs": [ + "mul_325", + "p_model_model_language_model_layers_23_mlp_down_proj_weight", + "p_model_model_language_model_layers_23_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_185", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl" + }, + { + "node": "add_257", + "inputs": [ + "add_254", + "linear_185" + ], + "output": "add_257", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl" + }, + { + "node": "mul_327", + "inputs": [ + "add_257", + "p_model_model_language_model_norm_weight" + ], + "output": "mul_327", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl" + }, + { + "node": "linear_186__part_0", + "inputs": [ + "mul_327", + "p_model_model_language_model_embed_tokens_weight__rows_0_131072", + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072" + ], + "output": "linear_186__part_0", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/c7c966b2f9c54a16d0078728edc1f2a7f79265fed388a815a0d52bcd73d1e0f6.wgsl" + }, + { + "node": "linear_186__part_1", + "inputs": [ + "mul_327", + "p_model_model_language_model_embed_tokens_weight__rows_131072_117248", + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248" + ], + "output": "linear_186__part_1", + "workgroups": [ + 1832, + 1, + 1 + ], + "kernel": "kernels/3fe93f80860ec8279a77e1dea55c00d6fe2bd81cb9d0c03624b80e9a999ec449.wgsl" + }, + { + "node": "linear_186", + "inputs": [ + "linear_186__part_0", + "linear_186__part_1" + ], + "output": "linear_186", + "workgroups": [ + 62080, + 1, + 1 + ], + "kernel": "kernels/026b53c4ecb0527a45fbd6020a725ad4d0ea2ea28c8cfa0a56e19cb0c1ad08ea.wgsl" + } + ], + "memory": { + "slots": [ + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 128, + 64, + 64, + 262144, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 134217728, + 16777216, + 120061952, + 15007744, + 65536, + 65536, + 65536, + 192, + 2048, + 2048, + 2048, + 2048, + 4096, + 393216, + 393216, + 491520, + 565248, + 131072, + 131072, + 1179648, + 16384, + 32768, + 8388608, + 7503872, + 15892480 + ], + "assignment": { + "p_model_model_language_model_layers_0_input_layernorm_weight": 0, + "p_model_model_language_model_layers_0_linear_attn_a_log": 1, + "p_model_model_language_model_layers_0_linear_attn_dt_bias": 2, + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight": 3, + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight": 4, + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight": 5, + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight": 6, + "p_model_model_language_model_layers_0_linear_attn_conv1d_weight": 7, + "p_model_model_language_model_layers_0_linear_attn_norm_weight": 8, + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight": 9, + "p_model_model_language_model_layers_0_post_attention_layernorm_weight": 10, + "p_model_model_language_model_layers_0_mlp_gate_proj_weight": 11, + "p_model_model_language_model_layers_0_mlp_up_proj_weight": 12, + "p_model_model_language_model_layers_0_mlp_down_proj_weight": 13, + "p_model_model_language_model_layers_1_input_layernorm_weight": 14, + "p_model_model_language_model_layers_1_linear_attn_a_log": 15, + "p_model_model_language_model_layers_1_linear_attn_dt_bias": 16, + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight": 17, + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight": 18, + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight": 19, + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight": 20, + "p_model_model_language_model_layers_1_linear_attn_conv1d_weight": 21, + "p_model_model_language_model_layers_1_linear_attn_norm_weight": 22, + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight": 23, + "p_model_model_language_model_layers_1_post_attention_layernorm_weight": 24, + "p_model_model_language_model_layers_1_mlp_gate_proj_weight": 25, + "p_model_model_language_model_layers_1_mlp_up_proj_weight": 26, + "p_model_model_language_model_layers_1_mlp_down_proj_weight": 27, + "p_model_model_language_model_layers_2_input_layernorm_weight": 28, + "p_model_model_language_model_layers_2_linear_attn_a_log": 29, + "p_model_model_language_model_layers_2_linear_attn_dt_bias": 30, + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight": 31, + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight": 32, + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight": 33, + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight": 34, + "p_model_model_language_model_layers_2_linear_attn_conv1d_weight": 35, + "p_model_model_language_model_layers_2_linear_attn_norm_weight": 36, + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight": 37, + "p_model_model_language_model_layers_2_post_attention_layernorm_weight": 38, + "p_model_model_language_model_layers_2_mlp_gate_proj_weight": 39, + "p_model_model_language_model_layers_2_mlp_up_proj_weight": 40, + "p_model_model_language_model_layers_2_mlp_down_proj_weight": 41, + "p_model_model_language_model_layers_3_input_layernorm_weight": 42, + "p_model_model_language_model_layers_3_self_attn_q_proj_weight": 43, + "p_model_model_language_model_layers_3_self_attn_q_norm_weight": 44, + "p_model_model_language_model_layers_3_self_attn_k_proj_weight": 45, + "p_model_model_language_model_layers_3_self_attn_k_norm_weight": 46, + "p_model_model_language_model_layers_3_self_attn_v_proj_weight": 47, + "p_model_model_language_model_layers_3_self_attn_o_proj_weight": 48, + "p_model_model_language_model_layers_3_post_attention_layernorm_weight": 49, + "p_model_model_language_model_layers_3_mlp_gate_proj_weight": 50, + "p_model_model_language_model_layers_3_mlp_up_proj_weight": 51, + "p_model_model_language_model_layers_3_mlp_down_proj_weight": 52, + "p_model_model_language_model_layers_4_input_layernorm_weight": 53, + "p_model_model_language_model_layers_4_linear_attn_a_log": 54, + "p_model_model_language_model_layers_4_linear_attn_dt_bias": 55, + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight": 56, + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight": 57, + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight": 58, + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight": 59, + "p_model_model_language_model_layers_4_linear_attn_conv1d_weight": 60, + "p_model_model_language_model_layers_4_linear_attn_norm_weight": 61, + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight": 62, + "p_model_model_language_model_layers_4_post_attention_layernorm_weight": 63, + "p_model_model_language_model_layers_4_mlp_gate_proj_weight": 64, + "p_model_model_language_model_layers_4_mlp_up_proj_weight": 65, + "p_model_model_language_model_layers_4_mlp_down_proj_weight": 66, + "p_model_model_language_model_layers_5_input_layernorm_weight": 67, + "p_model_model_language_model_layers_5_linear_attn_a_log": 68, + "p_model_model_language_model_layers_5_linear_attn_dt_bias": 69, + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight": 70, + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight": 71, + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight": 72, + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight": 73, + "p_model_model_language_model_layers_5_linear_attn_conv1d_weight": 74, + "p_model_model_language_model_layers_5_linear_attn_norm_weight": 75, + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight": 76, + "p_model_model_language_model_layers_5_post_attention_layernorm_weight": 77, + "p_model_model_language_model_layers_5_mlp_gate_proj_weight": 78, + "p_model_model_language_model_layers_5_mlp_up_proj_weight": 79, + "p_model_model_language_model_layers_5_mlp_down_proj_weight": 80, + "p_model_model_language_model_layers_6_input_layernorm_weight": 81, + "p_model_model_language_model_layers_6_linear_attn_a_log": 82, + "p_model_model_language_model_layers_6_linear_attn_dt_bias": 83, + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight": 84, + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight": 85, + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight": 86, + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight": 87, + "p_model_model_language_model_layers_6_linear_attn_conv1d_weight": 88, + "p_model_model_language_model_layers_6_linear_attn_norm_weight": 89, + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight": 90, + "p_model_model_language_model_layers_6_post_attention_layernorm_weight": 91, + "p_model_model_language_model_layers_6_mlp_gate_proj_weight": 92, + "p_model_model_language_model_layers_6_mlp_up_proj_weight": 93, + "p_model_model_language_model_layers_6_mlp_down_proj_weight": 94, + "p_model_model_language_model_layers_7_input_layernorm_weight": 95, + "p_model_model_language_model_layers_7_self_attn_q_proj_weight": 96, + "p_model_model_language_model_layers_7_self_attn_q_norm_weight": 97, + "p_model_model_language_model_layers_7_self_attn_k_proj_weight": 98, + "p_model_model_language_model_layers_7_self_attn_k_norm_weight": 99, + "p_model_model_language_model_layers_7_self_attn_v_proj_weight": 100, + "p_model_model_language_model_layers_7_self_attn_o_proj_weight": 101, + "p_model_model_language_model_layers_7_post_attention_layernorm_weight": 102, + "p_model_model_language_model_layers_7_mlp_gate_proj_weight": 103, + "p_model_model_language_model_layers_7_mlp_up_proj_weight": 104, + "p_model_model_language_model_layers_7_mlp_down_proj_weight": 105, + "p_model_model_language_model_layers_8_input_layernorm_weight": 106, + "p_model_model_language_model_layers_8_linear_attn_a_log": 107, + "p_model_model_language_model_layers_8_linear_attn_dt_bias": 108, + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight": 109, + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight": 110, + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight": 111, + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight": 112, + "p_model_model_language_model_layers_8_linear_attn_conv1d_weight": 113, + "p_model_model_language_model_layers_8_linear_attn_norm_weight": 114, + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight": 115, + "p_model_model_language_model_layers_8_post_attention_layernorm_weight": 116, + "p_model_model_language_model_layers_8_mlp_gate_proj_weight": 117, + "p_model_model_language_model_layers_8_mlp_up_proj_weight": 118, + "p_model_model_language_model_layers_8_mlp_down_proj_weight": 119, + "p_model_model_language_model_layers_9_input_layernorm_weight": 120, + "p_model_model_language_model_layers_9_linear_attn_a_log": 121, + "p_model_model_language_model_layers_9_linear_attn_dt_bias": 122, + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight": 123, + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight": 124, + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight": 125, + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight": 126, + "p_model_model_language_model_layers_9_linear_attn_conv1d_weight": 127, + "p_model_model_language_model_layers_9_linear_attn_norm_weight": 128, + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight": 129, + "p_model_model_language_model_layers_9_post_attention_layernorm_weight": 130, + "p_model_model_language_model_layers_9_mlp_gate_proj_weight": 131, + "p_model_model_language_model_layers_9_mlp_up_proj_weight": 132, + "p_model_model_language_model_layers_9_mlp_down_proj_weight": 133, + "p_model_model_language_model_layers_10_input_layernorm_weight": 134, + "p_model_model_language_model_layers_10_linear_attn_a_log": 135, + "p_model_model_language_model_layers_10_linear_attn_dt_bias": 136, + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight": 137, + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight": 138, + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight": 139, + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight": 140, + "p_model_model_language_model_layers_10_linear_attn_conv1d_weight": 141, + "p_model_model_language_model_layers_10_linear_attn_norm_weight": 142, + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight": 143, + "p_model_model_language_model_layers_10_post_attention_layernorm_weight": 144, + "p_model_model_language_model_layers_10_mlp_gate_proj_weight": 145, + "p_model_model_language_model_layers_10_mlp_up_proj_weight": 146, + "p_model_model_language_model_layers_10_mlp_down_proj_weight": 147, + "p_model_model_language_model_layers_11_input_layernorm_weight": 148, + "p_model_model_language_model_layers_11_self_attn_q_proj_weight": 149, + "p_model_model_language_model_layers_11_self_attn_q_norm_weight": 150, + "p_model_model_language_model_layers_11_self_attn_k_proj_weight": 151, + "p_model_model_language_model_layers_11_self_attn_k_norm_weight": 152, + "p_model_model_language_model_layers_11_self_attn_v_proj_weight": 153, + "p_model_model_language_model_layers_11_self_attn_o_proj_weight": 154, + "p_model_model_language_model_layers_11_post_attention_layernorm_weight": 155, + "p_model_model_language_model_layers_11_mlp_gate_proj_weight": 156, + "p_model_model_language_model_layers_11_mlp_up_proj_weight": 157, + "p_model_model_language_model_layers_11_mlp_down_proj_weight": 158, + "p_model_model_language_model_layers_12_input_layernorm_weight": 159, + "p_model_model_language_model_layers_12_linear_attn_a_log": 160, + "p_model_model_language_model_layers_12_linear_attn_dt_bias": 161, + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight": 162, + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight": 163, + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight": 164, + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight": 165, + "p_model_model_language_model_layers_12_linear_attn_conv1d_weight": 166, + "p_model_model_language_model_layers_12_linear_attn_norm_weight": 167, + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight": 168, + "p_model_model_language_model_layers_12_post_attention_layernorm_weight": 169, + "p_model_model_language_model_layers_12_mlp_gate_proj_weight": 170, + "p_model_model_language_model_layers_12_mlp_up_proj_weight": 171, + "p_model_model_language_model_layers_12_mlp_down_proj_weight": 172, + "p_model_model_language_model_layers_13_input_layernorm_weight": 173, + "p_model_model_language_model_layers_13_linear_attn_a_log": 174, + "p_model_model_language_model_layers_13_linear_attn_dt_bias": 175, + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight": 176, + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight": 177, + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight": 178, + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight": 179, + "p_model_model_language_model_layers_13_linear_attn_conv1d_weight": 180, + "p_model_model_language_model_layers_13_linear_attn_norm_weight": 181, + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight": 182, + "p_model_model_language_model_layers_13_post_attention_layernorm_weight": 183, + "p_model_model_language_model_layers_13_mlp_gate_proj_weight": 184, + "p_model_model_language_model_layers_13_mlp_up_proj_weight": 185, + "p_model_model_language_model_layers_13_mlp_down_proj_weight": 186, + "p_model_model_language_model_layers_14_input_layernorm_weight": 187, + "p_model_model_language_model_layers_14_linear_attn_a_log": 188, + "p_model_model_language_model_layers_14_linear_attn_dt_bias": 189, + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight": 190, + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight": 191, + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight": 192, + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight": 193, + "p_model_model_language_model_layers_14_linear_attn_conv1d_weight": 194, + "p_model_model_language_model_layers_14_linear_attn_norm_weight": 195, + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight": 196, + "p_model_model_language_model_layers_14_post_attention_layernorm_weight": 197, + "p_model_model_language_model_layers_14_mlp_gate_proj_weight": 198, + "p_model_model_language_model_layers_14_mlp_up_proj_weight": 199, + "p_model_model_language_model_layers_14_mlp_down_proj_weight": 200, + "p_model_model_language_model_layers_15_input_layernorm_weight": 201, + "p_model_model_language_model_layers_15_self_attn_q_proj_weight": 202, + "p_model_model_language_model_layers_15_self_attn_q_norm_weight": 203, + "p_model_model_language_model_layers_15_self_attn_k_proj_weight": 204, + "p_model_model_language_model_layers_15_self_attn_k_norm_weight": 205, + "p_model_model_language_model_layers_15_self_attn_v_proj_weight": 206, + "p_model_model_language_model_layers_15_self_attn_o_proj_weight": 207, + "p_model_model_language_model_layers_15_post_attention_layernorm_weight": 208, + "p_model_model_language_model_layers_15_mlp_gate_proj_weight": 209, + "p_model_model_language_model_layers_15_mlp_up_proj_weight": 210, + "p_model_model_language_model_layers_15_mlp_down_proj_weight": 211, + "p_model_model_language_model_layers_16_input_layernorm_weight": 212, + "p_model_model_language_model_layers_16_linear_attn_a_log": 213, + "p_model_model_language_model_layers_16_linear_attn_dt_bias": 214, + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight": 215, + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight": 216, + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight": 217, + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight": 218, + "p_model_model_language_model_layers_16_linear_attn_conv1d_weight": 219, + "p_model_model_language_model_layers_16_linear_attn_norm_weight": 220, + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight": 221, + "p_model_model_language_model_layers_16_post_attention_layernorm_weight": 222, + "p_model_model_language_model_layers_16_mlp_gate_proj_weight": 223, + "p_model_model_language_model_layers_16_mlp_up_proj_weight": 224, + "p_model_model_language_model_layers_16_mlp_down_proj_weight": 225, + "p_model_model_language_model_layers_17_input_layernorm_weight": 226, + "p_model_model_language_model_layers_17_linear_attn_a_log": 227, + "p_model_model_language_model_layers_17_linear_attn_dt_bias": 228, + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight": 229, + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight": 230, + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight": 231, + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight": 232, + "p_model_model_language_model_layers_17_linear_attn_conv1d_weight": 233, + "p_model_model_language_model_layers_17_linear_attn_norm_weight": 234, + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight": 235, + "p_model_model_language_model_layers_17_post_attention_layernorm_weight": 236, + "p_model_model_language_model_layers_17_mlp_gate_proj_weight": 237, + "p_model_model_language_model_layers_17_mlp_up_proj_weight": 238, + "p_model_model_language_model_layers_17_mlp_down_proj_weight": 239, + "p_model_model_language_model_layers_18_input_layernorm_weight": 240, + "p_model_model_language_model_layers_18_linear_attn_a_log": 241, + "p_model_model_language_model_layers_18_linear_attn_dt_bias": 242, + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight": 243, + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight": 244, + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight": 245, + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight": 246, + "p_model_model_language_model_layers_18_linear_attn_conv1d_weight": 247, + "p_model_model_language_model_layers_18_linear_attn_norm_weight": 248, + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight": 249, + "p_model_model_language_model_layers_18_post_attention_layernorm_weight": 250, + "p_model_model_language_model_layers_18_mlp_gate_proj_weight": 251, + "p_model_model_language_model_layers_18_mlp_up_proj_weight": 252, + "p_model_model_language_model_layers_18_mlp_down_proj_weight": 253, + "p_model_model_language_model_layers_19_input_layernorm_weight": 254, + "p_model_model_language_model_layers_19_self_attn_q_proj_weight": 255, + "p_model_model_language_model_layers_19_self_attn_q_norm_weight": 256, + "p_model_model_language_model_layers_19_self_attn_k_proj_weight": 257, + "p_model_model_language_model_layers_19_self_attn_k_norm_weight": 258, + "p_model_model_language_model_layers_19_self_attn_v_proj_weight": 259, + "p_model_model_language_model_layers_19_self_attn_o_proj_weight": 260, + "p_model_model_language_model_layers_19_post_attention_layernorm_weight": 261, + "p_model_model_language_model_layers_19_mlp_gate_proj_weight": 262, + "p_model_model_language_model_layers_19_mlp_up_proj_weight": 263, + "p_model_model_language_model_layers_19_mlp_down_proj_weight": 264, + "p_model_model_language_model_layers_20_input_layernorm_weight": 265, + "p_model_model_language_model_layers_20_linear_attn_a_log": 266, + "p_model_model_language_model_layers_20_linear_attn_dt_bias": 267, + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight": 268, + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight": 269, + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight": 270, + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight": 271, + "p_model_model_language_model_layers_20_linear_attn_conv1d_weight": 272, + "p_model_model_language_model_layers_20_linear_attn_norm_weight": 273, + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight": 274, + "p_model_model_language_model_layers_20_post_attention_layernorm_weight": 275, + "p_model_model_language_model_layers_20_mlp_gate_proj_weight": 276, + "p_model_model_language_model_layers_20_mlp_up_proj_weight": 277, + "p_model_model_language_model_layers_20_mlp_down_proj_weight": 278, + "p_model_model_language_model_layers_21_input_layernorm_weight": 279, + "p_model_model_language_model_layers_21_linear_attn_a_log": 280, + "p_model_model_language_model_layers_21_linear_attn_dt_bias": 281, + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight": 282, + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight": 283, + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight": 284, + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight": 285, + "p_model_model_language_model_layers_21_linear_attn_conv1d_weight": 286, + "p_model_model_language_model_layers_21_linear_attn_norm_weight": 287, + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight": 288, + "p_model_model_language_model_layers_21_post_attention_layernorm_weight": 289, + "p_model_model_language_model_layers_21_mlp_gate_proj_weight": 290, + "p_model_model_language_model_layers_21_mlp_up_proj_weight": 291, + "p_model_model_language_model_layers_21_mlp_down_proj_weight": 292, + "p_model_model_language_model_layers_22_input_layernorm_weight": 293, + "p_model_model_language_model_layers_22_linear_attn_a_log": 294, + "p_model_model_language_model_layers_22_linear_attn_dt_bias": 295, + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight": 296, + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight": 297, + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight": 298, + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight": 299, + "p_model_model_language_model_layers_22_linear_attn_conv1d_weight": 300, + "p_model_model_language_model_layers_22_linear_attn_norm_weight": 301, + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight": 302, + "p_model_model_language_model_layers_22_post_attention_layernorm_weight": 303, + "p_model_model_language_model_layers_22_mlp_gate_proj_weight": 304, + "p_model_model_language_model_layers_22_mlp_up_proj_weight": 305, + "p_model_model_language_model_layers_22_mlp_down_proj_weight": 306, + "p_model_model_language_model_layers_23_input_layernorm_weight": 307, + "p_model_model_language_model_layers_23_self_attn_q_proj_weight": 308, + "p_model_model_language_model_layers_23_self_attn_q_norm_weight": 309, + "p_model_model_language_model_layers_23_self_attn_k_proj_weight": 310, + "p_model_model_language_model_layers_23_self_attn_k_norm_weight": 311, + "p_model_model_language_model_layers_23_self_attn_v_proj_weight": 312, + "p_model_model_language_model_layers_23_self_attn_o_proj_weight": 313, + "p_model_model_language_model_layers_23_post_attention_layernorm_weight": 314, + "p_model_model_language_model_layers_23_mlp_gate_proj_weight": 315, + "p_model_model_language_model_layers_23_mlp_up_proj_weight": 316, + "p_model_model_language_model_layers_23_mlp_down_proj_weight": 317, + "p_model_model_language_model_norm_weight": 318, + "b_model_model_language_model_rotary_emb_inv_freq": 319, + "input_ids": 320, + "position_ids": 321, + "attention_mask": 322, + "states_0": 323, + "states_1": 324, + "states_2": 325, + "states_3": 326, + "states_4": 327, + "states_5": 328, + "states_6": 329, + "states_7": 330, + "states_8": 331, + "states_9": 332, + "states_10": 333, + "states_11": 334, + "states_12": 335, + "states_13": 336, + "states_14": 337, + "states_15": 338, + "states_16": 339, + "states_17": 340, + "states_18": 341, + "states_19": 342, + "states_20": 343, + "states_21": 344, + "states_22": 345, + "states_23": 346, + "states_24": 347, + "states_25": 348, + "states_26": 349, + "states_27": 350, + "states_28": 351, + "states_29": 352, + "states_30": 353, + "states_31": 354, + "states_32": 355, + "states_33": 356, + "states_34": 357, + "states_35": 358, + "states_36": 359, + "states_37": 360, + "states_38": 361, + "states_39": 362, + "states_40": 363, + "states_41": 364, + "states_42": 365, + "states_43": 366, + "states_44": 367, + "states_45": 368, + "states_46": 369, + "states_47": 370, + "p_model_model_language_model_layers_0_mlp_down_proj_weight__int4_scales": 371, + "p_model_model_language_model_layers_0_mlp_gate_proj_weight__int4_scales": 372, + "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales": 373, + "p_model_model_language_model_layers_10_mlp_down_proj_weight__int4_scales": 374, + "p_model_model_language_model_layers_10_mlp_gate_proj_weight__int4_scales": 375, + "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales": 376, + "p_model_model_language_model_layers_11_mlp_down_proj_weight__int4_scales": 377, + "p_model_model_language_model_layers_11_mlp_gate_proj_weight__int4_scales": 378, + "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales": 379, + "p_model_model_language_model_layers_12_mlp_down_proj_weight__int4_scales": 380, + "p_model_model_language_model_layers_12_mlp_gate_proj_weight__int4_scales": 381, + "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales": 382, + "p_model_model_language_model_layers_13_mlp_down_proj_weight__int4_scales": 383, + "p_model_model_language_model_layers_13_mlp_gate_proj_weight__int4_scales": 384, + "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales": 385, + "p_model_model_language_model_layers_14_mlp_down_proj_weight__int4_scales": 386, + "p_model_model_language_model_layers_14_mlp_gate_proj_weight__int4_scales": 387, + "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales": 388, + "p_model_model_language_model_layers_15_mlp_down_proj_weight__int4_scales": 389, + "p_model_model_language_model_layers_15_mlp_gate_proj_weight__int4_scales": 390, + "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales": 391, + "p_model_model_language_model_layers_16_mlp_down_proj_weight__int4_scales": 392, + "p_model_model_language_model_layers_16_mlp_gate_proj_weight__int4_scales": 393, + "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales": 394, + "p_model_model_language_model_layers_17_mlp_down_proj_weight__int4_scales": 395, + "p_model_model_language_model_layers_17_mlp_gate_proj_weight__int4_scales": 396, + "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales": 397, + "p_model_model_language_model_layers_18_mlp_down_proj_weight__int4_scales": 398, + "p_model_model_language_model_layers_18_mlp_gate_proj_weight__int4_scales": 399, + "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales": 400, + "p_model_model_language_model_layers_19_mlp_down_proj_weight__int4_scales": 401, + "p_model_model_language_model_layers_19_mlp_gate_proj_weight__int4_scales": 402, + "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales": 403, + "p_model_model_language_model_layers_1_mlp_down_proj_weight__int4_scales": 404, + "p_model_model_language_model_layers_1_mlp_gate_proj_weight__int4_scales": 405, + "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales": 406, + "p_model_model_language_model_layers_20_mlp_down_proj_weight__int4_scales": 407, + "p_model_model_language_model_layers_20_mlp_gate_proj_weight__int4_scales": 408, + "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales": 409, + "p_model_model_language_model_layers_21_mlp_down_proj_weight__int4_scales": 410, + "p_model_model_language_model_layers_21_mlp_gate_proj_weight__int4_scales": 411, + "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales": 412, + "p_model_model_language_model_layers_22_mlp_down_proj_weight__int4_scales": 413, + "p_model_model_language_model_layers_22_mlp_gate_proj_weight__int4_scales": 414, + "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales": 415, + "p_model_model_language_model_layers_23_mlp_down_proj_weight__int4_scales": 416, + "p_model_model_language_model_layers_23_mlp_gate_proj_weight__int4_scales": 417, + "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales": 418, + "p_model_model_language_model_layers_2_mlp_down_proj_weight__int4_scales": 419, + "p_model_model_language_model_layers_2_mlp_gate_proj_weight__int4_scales": 420, + "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales": 421, + "p_model_model_language_model_layers_3_mlp_down_proj_weight__int4_scales": 422, + "p_model_model_language_model_layers_3_mlp_gate_proj_weight__int4_scales": 423, + "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales": 424, + "p_model_model_language_model_layers_4_mlp_down_proj_weight__int4_scales": 425, + "p_model_model_language_model_layers_4_mlp_gate_proj_weight__int4_scales": 426, + "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales": 427, + "p_model_model_language_model_layers_5_mlp_down_proj_weight__int4_scales": 428, + "p_model_model_language_model_layers_5_mlp_gate_proj_weight__int4_scales": 429, + "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales": 430, + "p_model_model_language_model_layers_6_mlp_down_proj_weight__int4_scales": 431, + "p_model_model_language_model_layers_6_mlp_gate_proj_weight__int4_scales": 432, + "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales": 433, + "p_model_model_language_model_layers_7_mlp_down_proj_weight__int4_scales": 434, + "p_model_model_language_model_layers_7_mlp_gate_proj_weight__int4_scales": 435, + "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales": 436, + "p_model_model_language_model_layers_8_mlp_down_proj_weight__int4_scales": 437, + "p_model_model_language_model_layers_8_mlp_gate_proj_weight__int4_scales": 438, + "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales": 439, + "p_model_model_language_model_layers_9_mlp_down_proj_weight__int4_scales": 440, + "p_model_model_language_model_layers_9_mlp_gate_proj_weight__int4_scales": 441, + "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales": 442, + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight__int8_scales": 443, + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight__int8_scales": 444, + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight__int8_scales": 445, + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight__int8_scales": 446, + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight__int8_scales": 447, + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight__int8_scales": 448, + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight__int8_scales": 449, + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight__int8_scales": 450, + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight__int8_scales": 451, + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight__int8_scales": 452, + "p_model_model_language_model_layers_11_self_attn_k_proj_weight__int8_scales": 453, + "p_model_model_language_model_layers_11_self_attn_o_proj_weight__int8_scales": 454, + "p_model_model_language_model_layers_11_self_attn_q_proj_weight__int8_scales": 455, + "p_model_model_language_model_layers_11_self_attn_v_proj_weight__int8_scales": 456, + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight__int8_scales": 457, + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight__int8_scales": 458, + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight__int8_scales": 459, + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight__int8_scales": 460, + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight__int8_scales": 461, + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight__int8_scales": 462, + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight__int8_scales": 463, + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight__int8_scales": 464, + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight__int8_scales": 465, + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight__int8_scales": 466, + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight__int8_scales": 467, + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight__int8_scales": 468, + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight__int8_scales": 469, + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight__int8_scales": 470, + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight__int8_scales": 471, + "p_model_model_language_model_layers_15_self_attn_k_proj_weight__int8_scales": 472, + "p_model_model_language_model_layers_15_self_attn_o_proj_weight__int8_scales": 473, + "p_model_model_language_model_layers_15_self_attn_q_proj_weight__int8_scales": 474, + "p_model_model_language_model_layers_15_self_attn_v_proj_weight__int8_scales": 475, + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight__int8_scales": 476, + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight__int8_scales": 477, + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight__int8_scales": 478, + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight__int8_scales": 479, + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight__int8_scales": 480, + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight__int8_scales": 481, + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight__int8_scales": 482, + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight__int8_scales": 483, + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight__int8_scales": 484, + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight__int8_scales": 485, + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight__int8_scales": 486, + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight__int8_scales": 487, + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight__int8_scales": 488, + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight__int8_scales": 489, + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight__int8_scales": 490, + "p_model_model_language_model_layers_19_self_attn_k_proj_weight__int8_scales": 491, + "p_model_model_language_model_layers_19_self_attn_o_proj_weight__int8_scales": 492, + "p_model_model_language_model_layers_19_self_attn_q_proj_weight__int8_scales": 493, + "p_model_model_language_model_layers_19_self_attn_v_proj_weight__int8_scales": 494, + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight__int8_scales": 495, + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight__int8_scales": 496, + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight__int8_scales": 497, + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight__int8_scales": 498, + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight__int8_scales": 499, + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight__int8_scales": 500, + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight__int8_scales": 501, + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight__int8_scales": 502, + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight__int8_scales": 503, + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight__int8_scales": 504, + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight__int8_scales": 505, + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight__int8_scales": 506, + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight__int8_scales": 507, + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight__int8_scales": 508, + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight__int8_scales": 509, + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight__int8_scales": 510, + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight__int8_scales": 511, + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight__int8_scales": 512, + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight__int8_scales": 513, + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight__int8_scales": 514, + "p_model_model_language_model_layers_23_self_attn_k_proj_weight__int8_scales": 515, + "p_model_model_language_model_layers_23_self_attn_o_proj_weight__int8_scales": 516, + "p_model_model_language_model_layers_23_self_attn_q_proj_weight__int8_scales": 517, + "p_model_model_language_model_layers_23_self_attn_v_proj_weight__int8_scales": 518, + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight__int8_scales": 519, + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight__int8_scales": 520, + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight__int8_scales": 521, + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight__int8_scales": 522, + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight__int8_scales": 523, + "p_model_model_language_model_layers_3_self_attn_k_proj_weight__int8_scales": 524, + "p_model_model_language_model_layers_3_self_attn_o_proj_weight__int8_scales": 525, + "p_model_model_language_model_layers_3_self_attn_q_proj_weight__int8_scales": 526, + "p_model_model_language_model_layers_3_self_attn_v_proj_weight__int8_scales": 527, + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight__int8_scales": 528, + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight__int8_scales": 529, + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight__int8_scales": 530, + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight__int8_scales": 531, + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight__int8_scales": 532, + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight__int8_scales": 533, + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight__int8_scales": 534, + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight__int8_scales": 535, + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight__int8_scales": 536, + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight__int8_scales": 537, + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight__int8_scales": 538, + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight__int8_scales": 539, + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight__int8_scales": 540, + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight__int8_scales": 541, + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight__int8_scales": 542, + "p_model_model_language_model_layers_7_self_attn_k_proj_weight__int8_scales": 543, + "p_model_model_language_model_layers_7_self_attn_o_proj_weight__int8_scales": 544, + "p_model_model_language_model_layers_7_self_attn_q_proj_weight__int8_scales": 545, + "p_model_model_language_model_layers_7_self_attn_v_proj_weight__int8_scales": 546, + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight__int8_scales": 547, + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight__int8_scales": 548, + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight__int8_scales": 549, + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight__int8_scales": 550, + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight__int8_scales": 551, + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight__int8_scales": 552, + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight__int8_scales": 553, + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight__int8_scales": 554, + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight__int8_scales": 555, + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight__int8_scales": 556, + "p_model_model_language_model_embed_tokens_weight__rows_0_131072": 557, + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072": 558, + "p_model_model_language_model_embed_tokens_weight__rows_131072_117248": 559, + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248": 560, + "embedding__part_0": 561, + "embedding__part_1": 562, + "embedding": 563, + "unsqueeze": 561, + "expand": 562, + "slice_1": 561, + "expand_1": 562, + "_to_copy": 564, + "matmul": 561, + "transpose": 562, + "select": 565, + "select_1": 566, + "slice_2": 562, + "slice_3": 566, + "copy": 565, + "transpose_1": 562, + "select_2": 561, + "slice_scatter": 567, + "select_scatter": 561, + "transpose_2": 562, + "transpose_7": 561, + "select_7": 565, + "slice_7": 567, + "transpose_8": 561, + "select_8": 565, + "slice_8": 561, + "copy_1": 565, + "transpose_9": 561, + "select_9": 562, + "slice_scatter_1": 568, + "select_scatter_1": 562, + "transpose_10": 561, + "transpose_12": 562, + "select_11": 565, + "cat": 561, + "cos": 562, + "mul": 569, + "sin": 562, + "mul_1": 561, + "mul_3": 562, + "linear": 570, + "transpose_13": 571, + "linear_1": 570, + "linear_2": 568, + "linear_3": 565, + "cat_1": 572, + "slice_10": 323, + "conv1d": 573, + "slice_11": 572, + "silu": 573, + "slice_12": 571, + "transpose_14": 572, + "slice_13": 571, + "slice_14": 573, + "slice_15": 574, + "sigmoid": 562, + "exp": 564, + "neg": 568, + "add_2": 572, + "softplus": 565, + "mul_4": 572, + "mul_5": 575, + "sum_1": 568, + "add_3": 565, + "rsqrt_1": 568, + "mul_6": 575, + "mul_7": 571, + "sum_2": 565, + "add_4": 568, + "rsqrt_2": 565, + "mul_8": 571, + "normalized_gated_delta_rule": 576, + "slice_16": 324, + "slice_17": 575, + "transpose_15": 574, + "mul_10": 575, + "mul_11": 574, + "linear_4": 562, + "add_6": 575, + "mul_13": 563, + "linear_5": 571, + "linear_6": 570, + "mul_14": 572, + "linear_7": 562, + "add_9": 563, + "mul_16": 562, + "linear_8": 571, + "transpose_16": 570, + "linear_9": 574, + "linear_10": 568, + "linear_11": 565, + "cat_2": 572, + "slice_18": 325, + "conv1d_1": 573, + "slice_19": 572, + "silu_3": 573, + "slice_20": 571, + "transpose_17": 570, + "slice_21": 575, + "slice_22": 571, + "slice_23": 572, + "sigmoid_1": 562, + "exp_1": 564, + "neg_1": 568, + "add_12": 570, + "softplus_1": 565, + "mul_17": 570, + "mul_18": 573, + "sum_3": 568, + "add_13": 565, + "rsqrt_6": 568, + "mul_19": 573, + "mul_20": 575, + "sum_4": 565, + "add_14": 568, + "rsqrt_7": 565, + "mul_21": 575, + "normalized_gated_delta_rule_1": 576, + "slice_24": 326, + "slice_25": 575, + "transpose_18": 571, + "mul_23": 575, + "mul_24": 570, + "linear_12": 562, + "add_16": 574, + "mul_26": 563, + "linear_13": 571, + "linear_14": 570, + "mul_27": 572, + "linear_15": 562, + "add_19": 563, + "mul_29": 562, + "linear_16": 571, + "transpose_19": 570, + "linear_17": 575, + "linear_18": 568, + "linear_19": 565, + "cat_3": 572, + "slice_26": 327, + "conv1d_2": 573, + "slice_27": 572, + "silu_6": 573, + "slice_28": 571, + "transpose_20": 570, + "slice_29": 574, + "slice_30": 571, + "slice_31": 572, + "sigmoid_2": 562, + "exp_2": 564, + "neg_2": 568, + "add_22": 570, + "softplus_2": 565, + "mul_30": 570, + "mul_31": 573, + "sum_5": 568, + "add_23": 565, + "rsqrt_11": 568, + "mul_32": 573, + "mul_33": 574, + "sum_6": 565, + "add_24": 568, + "rsqrt_12": 565, + "mul_34": 574, + "normalized_gated_delta_rule_2": 576, + "slice_32": 328, + "slice_33": 574, + "transpose_21": 571, + "mul_36": 574, + "mul_37": 570, + "linear_20": 562, + "add_26": 575, + "mul_39": 563, + "linear_21": 571, + "linear_22": 570, + "mul_40": 572, + "linear_23": 562, + "add_29": 563, + "mul_42": 562, + "linear_24": 571, + "slice_34": 574, + "slice_35": 575, + "mul_44": 570, + "transpose_22": 574, + "linear_25": 571, + "mul_46": 570, + "transpose_23": 571, + "linear_26": 570, + "transpose_24": 562, + "slice_36": 570, + "slice_37": 572, + "slice_38": 574, + "slice_39": 573, + "mul_47": 571, + "slice_40": 576, + "slice_41": 577, + "neg_3": 570, + "cat_4": 578, + "mul_48": 570, + "add_36": 578, + "mul_49": 577, + "slice_42": 571, + "slice_43": 570, + "neg_4": 574, + "cat_5": 570, + "mul_50": 574, + "add_37": 571, + "cat_6": 574, + "cat_7": 578, + "cache_update": 329, + "cache_update_1": 330, + "matmul_2": 570, + "transpose_26": 574, + "sigmoid_3": 571, + "mul_52": 575, + "linear_27": 562, + "add_39": 574, + "mul_54": 563, + "linear_28": 570, + "linear_29": 571, + "mul_55": 572, + "linear_30": 562, + "add_42": 563, + "mul_57": 562, + "linear_31": 570, + "transpose_27": 571, + "linear_32": 575, + "linear_33": 568, + "linear_34": 565, + "cat_8": 572, + "slice_44": 331, + "conv1d_3": 573, + "slice_45": 572, + "silu_10": 573, + "slice_46": 570, + "transpose_28": 571, + "slice_47": 574, + "slice_48": 570, + "slice_49": 572, + "sigmoid_4": 577, + "exp_3": 564, + "neg_5": 568, + "add_45": 578, + "softplus_3": 565, + "mul_58": 578, + "mul_59": 571, + "sum_7": 568, + "add_46": 565, + "rsqrt_20": 568, + "mul_60": 571, + "mul_61": 574, + "sum_8": 565, + "add_47": 568, + "rsqrt_21": 565, + "mul_62": 574, + "normalized_gated_delta_rule_3": 576, + "slice_50": 332, + "slice_51": 574, + "transpose_29": 570, + "mul_64": 574, + "mul_65": 571, + "linear_35": 562, + "add_49": 575, + "mul_67": 563, + "linear_36": 570, + "linear_37": 571, + "mul_68": 572, + "linear_38": 562, + "add_52": 563, + "mul_70": 562, + "linear_39": 570, + "transpose_30": 571, + "linear_40": 574, + "linear_41": 568, + "linear_42": 565, + "cat_9": 572, + "slice_52": 333, + "conv1d_4": 573, + "slice_53": 572, + "silu_13": 573, + "slice_54": 570, + "transpose_31": 571, + "slice_55": 575, + "slice_56": 570, + "slice_57": 572, + "sigmoid_5": 577, + "exp_4": 564, + "neg_6": 568, + "add_55": 578, + "softplus_4": 565, + "mul_71": 578, + "mul_72": 571, + "sum_9": 568, + "add_56": 565, + "rsqrt_25": 568, + "mul_73": 571, + "mul_74": 575, + "sum_10": 565, + "add_57": 568, + "rsqrt_26": 565, + "mul_75": 575, + "normalized_gated_delta_rule_4": 576, + "slice_58": 334, + "slice_59": 575, + "transpose_32": 570, + "mul_77": 575, + "mul_78": 571, + "linear_43": 562, + "add_59": 574, + "mul_80": 563, + "linear_44": 570, + "linear_45": 571, + "mul_81": 572, + "linear_46": 562, + "add_62": 563, + "mul_83": 562, + "linear_47": 570, + "transpose_33": 571, + "linear_48": 575, + "linear_49": 568, + "linear_50": 565, + "cat_10": 572, + "slice_60": 335, + "conv1d_5": 573, + "slice_61": 572, + "silu_16": 573, + "slice_62": 570, + "transpose_34": 571, + "slice_63": 574, + "slice_64": 570, + "slice_65": 572, + "sigmoid_6": 577, + "exp_5": 564, + "neg_7": 568, + "add_65": 578, + "softplus_5": 565, + "mul_84": 578, + "mul_85": 571, + "sum_11": 568, + "add_66": 565, + "rsqrt_30": 568, + "mul_86": 571, + "mul_87": 574, + "sum_12": 565, + "add_67": 568, + "rsqrt_31": 565, + "mul_88": 574, + "normalized_gated_delta_rule_5": 576, + "slice_66": 336, + "slice_67": 574, + "transpose_35": 570, + "mul_90": 574, + "mul_91": 571, + "linear_51": 562, + "add_69": 575, + "mul_93": 563, + "linear_52": 570, + "linear_53": 571, + "mul_94": 572, + "linear_54": 562, + "add_72": 563, + "mul_96": 562, + "linear_55": 570, + "slice_68": 574, + "slice_69": 575, + "mul_98": 571, + "transpose_36": 574, + "linear_56": 578, + "mul_100": 570, + "transpose_37": 578, + "linear_57": 571, + "transpose_38": 562, + "slice_70": 570, + "slice_71": 571, + "slice_72": 577, + "slice_73": 574, + "mul_101": 578, + "slice_74": 572, + "slice_75": 573, + "neg_8": 570, + "cat_11": 573, + "mul_102": 570, + "add_79": 572, + "mul_103": 578, + "slice_76": 570, + "slice_77": 573, + "neg_9": 577, + "cat_12": 573, + "mul_104": 577, + "add_80": 570, + "cat_13": 573, + "cat_14": 578, + "cache_update_2": 337, + "cache_update_3": 338, + "matmul_4": 574, + "transpose_40": 571, + "sigmoid_7": 574, + "mul_106": 575, + "linear_58": 562, + "add_82": 574, + "mul_108": 563, + "linear_59": 570, + "linear_60": 571, + "mul_109": 572, + "linear_61": 562, + "add_85": 563, + "mul_111": 562, + "linear_62": 570, + "transpose_41": 571, + "linear_63": 575, + "linear_64": 568, + "linear_65": 565, + "cat_15": 572, + "slice_78": 339, + "conv1d_6": 573, + "slice_79": 572, + "silu_20": 573, + "slice_80": 570, + "transpose_42": 571, + "slice_81": 574, + "slice_82": 570, + "slice_83": 572, + "sigmoid_8": 577, + "exp_6": 564, + "neg_10": 568, + "add_88": 578, + "softplus_6": 565, + "mul_112": 578, + "mul_113": 571, + "sum_13": 568, + "add_89": 565, + "rsqrt_39": 568, + "mul_114": 571, + "mul_115": 574, + "sum_14": 565, + "add_90": 568, + "rsqrt_40": 565, + "mul_116": 574, + "normalized_gated_delta_rule_6": 576, + "slice_84": 340, + "slice_85": 574, + "transpose_43": 570, + "mul_118": 574, + "mul_119": 571, + "linear_66": 562, + "add_92": 575, + "mul_121": 563, + "linear_67": 570, + "linear_68": 571, + "mul_122": 572, + "linear_69": 562, + "add_95": 563, + "mul_124": 562, + "linear_70": 570, + "transpose_44": 571, + "linear_71": 574, + "linear_72": 568, + "linear_73": 565, + "cat_16": 572, + "slice_86": 341, + "conv1d_7": 573, + "slice_87": 572, + "silu_23": 573, + "slice_88": 570, + "transpose_45": 571, + "slice_89": 575, + "slice_90": 570, + "slice_91": 572, + "sigmoid_9": 577, + "exp_7": 564, + "neg_11": 568, + "add_98": 578, + "softplus_7": 565, + "mul_125": 578, + "mul_126": 571, + "sum_15": 568, + "add_99": 565, + "rsqrt_44": 568, + "mul_127": 571, + "mul_128": 575, + "sum_16": 565, + "add_100": 568, + "rsqrt_45": 565, + "mul_129": 575, + "normalized_gated_delta_rule_7": 576, + "slice_92": 342, + "slice_93": 575, + "transpose_46": 570, + "mul_131": 575, + "mul_132": 571, + "linear_74": 562, + "add_102": 574, + "mul_134": 563, + "linear_75": 570, + "linear_76": 571, + "mul_135": 572, + "linear_77": 562, + "add_105": 563, + "mul_137": 562, + "linear_78": 570, + "transpose_47": 571, + "linear_79": 575, + "linear_80": 568, + "linear_81": 565, + "cat_17": 572, + "slice_94": 343, + "conv1d_8": 573, + "slice_95": 572, + "silu_26": 573, + "slice_96": 570, + "transpose_48": 571, + "slice_97": 574, + "slice_98": 570, + "slice_99": 572, + "sigmoid_10": 577, + "exp_8": 564, + "neg_12": 568, + "add_108": 578, + "softplus_8": 565, + "mul_138": 578, + "mul_139": 571, + "sum_17": 568, + "add_109": 565, + "rsqrt_49": 568, + "mul_140": 571, + "mul_141": 574, + "sum_18": 565, + "add_110": 568, + "rsqrt_50": 565, + "mul_142": 574, + "normalized_gated_delta_rule_8": 576, + "slice_100": 344, + "slice_101": 574, + "transpose_49": 570, + "mul_144": 574, + "mul_145": 571, + "linear_82": 562, + "add_112": 575, + "mul_147": 563, + "linear_83": 570, + "linear_84": 571, + "mul_148": 572, + "linear_85": 562, + "add_115": 563, + "mul_150": 562, + "linear_86": 570, + "slice_102": 574, + "slice_103": 575, + "mul_152": 571, + "transpose_50": 574, + "linear_87": 578, + "mul_154": 570, + "transpose_51": 578, + "linear_88": 571, + "transpose_52": 562, + "slice_104": 570, + "slice_105": 571, + "slice_106": 577, + "slice_107": 574, + "mul_155": 578, + "slice_108": 572, + "slice_109": 573, + "neg_13": 570, + "cat_18": 573, + "mul_156": 570, + "add_122": 572, + "mul_157": 578, + "slice_110": 570, + "slice_111": 573, + "neg_14": 577, + "cat_19": 573, + "mul_158": 577, + "add_123": 570, + "cat_20": 573, + "cat_21": 578, + "cache_update_4": 345, + "cache_update_5": 346, + "matmul_6": 574, + "transpose_54": 571, + "sigmoid_11": 574, + "mul_160": 575, + "linear_89": 562, + "add_125": 574, + "mul_162": 563, + "linear_90": 570, + "linear_91": 571, + "mul_163": 572, + "linear_92": 562, + "add_128": 563, + "mul_165": 562, + "linear_93": 570, + "transpose_55": 571, + "linear_94": 575, + "linear_95": 568, + "linear_96": 565, + "cat_22": 572, + "slice_112": 347, + "conv1d_9": 573, + "slice_113": 572, + "silu_30": 573, + "slice_114": 570, + "transpose_56": 571, + "slice_115": 574, + "slice_116": 570, + "slice_117": 572, + "sigmoid_12": 577, + "exp_9": 564, + "neg_15": 568, + "add_131": 578, + "softplus_9": 565, + "mul_166": 578, + "mul_167": 571, + "sum_19": 568, + "add_132": 565, + "rsqrt_58": 568, + "mul_168": 571, + "mul_169": 574, + "sum_20": 565, + "add_133": 568, + "rsqrt_59": 565, + "mul_170": 574, + "normalized_gated_delta_rule_9": 576, + "slice_118": 348, + "slice_119": 574, + "transpose_57": 570, + "mul_172": 574, + "mul_173": 571, + "linear_97": 562, + "add_135": 575, + "mul_175": 563, + "linear_98": 570, + "linear_99": 571, + "mul_176": 572, + "linear_100": 562, + "add_138": 563, + "mul_178": 562, + "linear_101": 570, + "transpose_58": 571, + "linear_102": 574, + "linear_103": 568, + "linear_104": 565, + "cat_23": 572, + "slice_120": 349, + "conv1d_10": 573, + "slice_121": 572, + "silu_33": 573, + "slice_122": 570, + "transpose_59": 571, + "slice_123": 575, + "slice_124": 570, + "slice_125": 572, + "sigmoid_13": 577, + "exp_10": 564, + "neg_16": 568, + "add_141": 578, + "softplus_10": 565, + "mul_179": 578, + "mul_180": 571, + "sum_21": 568, + "add_142": 565, + "rsqrt_63": 568, + "mul_181": 571, + "mul_182": 575, + "sum_22": 565, + "add_143": 568, + "rsqrt_64": 565, + "mul_183": 575, + "normalized_gated_delta_rule_10": 576, + "slice_126": 350, + "slice_127": 575, + "transpose_60": 570, + "mul_185": 575, + "mul_186": 571, + "linear_105": 562, + "add_145": 574, + "mul_188": 563, + "linear_106": 570, + "linear_107": 571, + "mul_189": 572, + "linear_108": 562, + "add_148": 563, + "mul_191": 562, + "linear_109": 570, + "transpose_61": 571, + "linear_110": 575, + "linear_111": 568, + "linear_112": 565, + "cat_24": 572, + "slice_128": 351, + "conv1d_11": 573, + "slice_129": 572, + "silu_36": 573, + "slice_130": 570, + "transpose_62": 571, + "slice_131": 574, + "slice_132": 570, + "slice_133": 572, + "sigmoid_14": 577, + "exp_11": 564, + "neg_17": 568, + "add_151": 578, + "softplus_11": 565, + "mul_192": 578, + "mul_193": 571, + "sum_23": 568, + "add_152": 565, + "rsqrt_68": 568, + "mul_194": 571, + "mul_195": 574, + "sum_24": 565, + "add_153": 568, + "rsqrt_69": 565, + "mul_196": 574, + "normalized_gated_delta_rule_11": 576, + "slice_134": 352, + "slice_135": 574, + "transpose_63": 570, + "mul_198": 574, + "mul_199": 571, + "linear_113": 562, + "add_155": 575, + "mul_201": 563, + "linear_114": 570, + "linear_115": 571, + "mul_202": 572, + "linear_116": 562, + "add_158": 563, + "mul_204": 562, + "linear_117": 570, + "slice_136": 574, + "slice_137": 575, + "mul_206": 571, + "transpose_64": 574, + "linear_118": 578, + "mul_208": 570, + "transpose_65": 578, + "linear_119": 571, + "transpose_66": 562, + "slice_138": 570, + "slice_139": 571, + "slice_140": 577, + "slice_141": 574, + "mul_209": 578, + "slice_142": 572, + "slice_143": 573, + "neg_18": 570, + "cat_25": 573, + "mul_210": 570, + "add_165": 572, + "mul_211": 578, + "slice_144": 570, + "slice_145": 573, + "neg_19": 577, + "cat_26": 573, + "mul_212": 577, + "add_166": 570, + "cat_27": 573, + "cat_28": 578, + "cache_update_6": 353, + "cache_update_7": 354, + "matmul_8": 574, + "transpose_68": 571, + "sigmoid_15": 574, + "mul_214": 575, + "linear_120": 562, + "add_168": 574, + "mul_216": 563, + "linear_121": 570, + "linear_122": 571, + "mul_217": 572, + "linear_123": 562, + "add_171": 563, + "mul_219": 562, + "linear_124": 570, + "transpose_69": 571, + "linear_125": 575, + "linear_126": 568, + "linear_127": 565, + "cat_29": 572, + "slice_146": 355, + "conv1d_12": 573, + "slice_147": 572, + "silu_40": 573, + "slice_148": 570, + "transpose_70": 571, + "slice_149": 574, + "slice_150": 570, + "slice_151": 572, + "sigmoid_16": 577, + "exp_12": 564, + "neg_20": 568, + "add_174": 578, + "softplus_12": 565, + "mul_220": 578, + "mul_221": 571, + "sum_25": 568, + "add_175": 565, + "rsqrt_77": 568, + "mul_222": 571, + "mul_223": 574, + "sum_26": 565, + "add_176": 568, + "rsqrt_78": 565, + "mul_224": 574, + "normalized_gated_delta_rule_12": 576, + "slice_152": 356, + "slice_153": 574, + "transpose_71": 570, + "mul_226": 574, + "mul_227": 571, + "linear_128": 562, + "add_178": 575, + "mul_229": 563, + "linear_129": 570, + "linear_130": 571, + "mul_230": 572, + "linear_131": 562, + "add_181": 563, + "mul_232": 562, + "linear_132": 570, + "transpose_72": 571, + "linear_133": 574, + "linear_134": 568, + "linear_135": 565, + "cat_30": 572, + "slice_154": 357, + "conv1d_13": 573, + "slice_155": 572, + "silu_43": 573, + "slice_156": 570, + "transpose_73": 571, + "slice_157": 575, + "slice_158": 570, + "slice_159": 572, + "sigmoid_17": 577, + "exp_13": 564, + "neg_21": 568, + "add_184": 578, + "softplus_13": 565, + "mul_233": 578, + "mul_234": 571, + "sum_27": 568, + "add_185": 565, + "rsqrt_82": 568, + "mul_235": 571, + "mul_236": 575, + "sum_28": 565, + "add_186": 568, + "rsqrt_83": 565, + "mul_237": 575, + "normalized_gated_delta_rule_13": 576, + "slice_160": 358, + "slice_161": 575, + "transpose_74": 570, + "mul_239": 575, + "mul_240": 571, + "linear_136": 562, + "add_188": 574, + "mul_242": 563, + "linear_137": 570, + "linear_138": 571, + "mul_243": 572, + "linear_139": 562, + "add_191": 563, + "mul_245": 562, + "linear_140": 570, + "transpose_75": 571, + "linear_141": 575, + "linear_142": 568, + "linear_143": 565, + "cat_31": 572, + "slice_162": 359, + "conv1d_14": 573, + "slice_163": 572, + "silu_46": 573, + "slice_164": 570, + "transpose_76": 571, + "slice_165": 574, + "slice_166": 570, + "slice_167": 572, + "sigmoid_18": 577, + "exp_14": 564, + "neg_22": 568, + "add_194": 578, + "softplus_14": 565, + "mul_246": 578, + "mul_247": 571, + "sum_29": 568, + "add_195": 565, + "rsqrt_87": 568, + "mul_248": 571, + "mul_249": 574, + "sum_30": 565, + "add_196": 568, + "rsqrt_88": 565, + "mul_250": 574, + "normalized_gated_delta_rule_14": 576, + "slice_168": 360, + "slice_169": 574, + "transpose_77": 570, + "mul_252": 574, + "mul_253": 571, + "linear_144": 562, + "add_198": 575, + "mul_255": 563, + "linear_145": 570, + "linear_146": 571, + "mul_256": 572, + "linear_147": 562, + "add_201": 563, + "mul_258": 562, + "linear_148": 570, + "slice_170": 574, + "slice_171": 575, + "mul_260": 571, + "transpose_78": 574, + "linear_149": 578, + "mul_262": 570, + "transpose_79": 578, + "linear_150": 571, + "transpose_80": 562, + "slice_172": 570, + "slice_173": 571, + "slice_174": 577, + "slice_175": 574, + "mul_263": 578, + "slice_176": 572, + "slice_177": 573, + "neg_23": 570, + "cat_32": 573, + "mul_264": 570, + "add_208": 572, + "mul_265": 578, + "slice_178": 570, + "slice_179": 573, + "neg_24": 577, + "cat_33": 573, + "mul_266": 577, + "add_209": 570, + "cat_34": 573, + "cat_35": 578, + "cache_update_8": 361, + "cache_update_9": 362, + "matmul_10": 574, + "transpose_82": 571, + "sigmoid_19": 574, + "mul_268": 575, + "linear_151": 562, + "add_211": 574, + "mul_270": 563, + "linear_152": 570, + "linear_153": 571, + "mul_271": 572, + "linear_154": 562, + "add_214": 563, + "mul_273": 562, + "linear_155": 570, + "transpose_83": 571, + "linear_156": 575, + "linear_157": 568, + "linear_158": 565, + "cat_36": 572, + "slice_180": 363, + "conv1d_15": 573, + "slice_181": 572, + "silu_50": 573, + "slice_182": 570, + "transpose_84": 571, + "slice_183": 574, + "slice_184": 570, + "slice_185": 572, + "sigmoid_20": 577, + "exp_15": 564, + "neg_25": 568, + "add_217": 578, + "softplus_15": 565, + "mul_274": 578, + "mul_275": 571, + "sum_31": 568, + "add_218": 565, + "rsqrt_96": 568, + "mul_276": 571, + "mul_277": 574, + "sum_32": 565, + "add_219": 568, + "rsqrt_97": 565, + "mul_278": 574, + "normalized_gated_delta_rule_15": 576, + "slice_186": 364, + "slice_187": 574, + "transpose_85": 570, + "mul_280": 574, + "mul_281": 571, + "linear_159": 562, + "add_221": 575, + "mul_283": 563, + "linear_160": 570, + "linear_161": 571, + "mul_284": 572, + "linear_162": 562, + "add_224": 563, + "mul_286": 562, + "linear_163": 570, + "transpose_86": 571, + "linear_164": 574, + "linear_165": 568, + "linear_166": 565, + "cat_37": 572, + "slice_188": 365, + "conv1d_16": 573, + "slice_189": 572, + "silu_53": 573, + "slice_190": 570, + "transpose_87": 571, + "slice_191": 575, + "slice_192": 570, + "slice_193": 572, + "sigmoid_21": 577, + "exp_16": 564, + "neg_26": 568, + "add_227": 578, + "softplus_16": 565, + "mul_287": 578, + "mul_288": 571, + "sum_33": 568, + "add_228": 565, + "rsqrt_101": 568, + "mul_289": 571, + "mul_290": 575, + "sum_34": 565, + "add_229": 568, + "rsqrt_102": 565, + "mul_291": 575, + "normalized_gated_delta_rule_16": 576, + "slice_194": 366, + "slice_195": 575, + "transpose_88": 570, + "mul_293": 575, + "mul_294": 571, + "linear_167": 562, + "add_231": 574, + "mul_296": 563, + "linear_168": 570, + "linear_169": 571, + "mul_297": 572, + "linear_170": 562, + "add_234": 563, + "mul_299": 562, + "linear_171": 570, + "transpose_89": 571, + "linear_172": 575, + "linear_173": 568, + "linear_174": 565, + "cat_38": 572, + "slice_196": 367, + "conv1d_17": 573, + "slice_197": 572, + "silu_56": 573, + "slice_198": 570, + "transpose_90": 571, + "slice_199": 574, + "slice_200": 570, + "slice_201": 572, + "sigmoid_22": 577, + "exp_17": 564, + "neg_27": 568, + "add_237": 578, + "softplus_17": 565, + "mul_300": 578, + "mul_301": 571, + "sum_35": 568, + "add_238": 565, + "rsqrt_106": 568, + "mul_302": 571, + "mul_303": 574, + "sum_36": 565, + "add_239": 568, + "rsqrt_107": 565, + "mul_304": 574, + "normalized_gated_delta_rule_17": 576, + "slice_202": 368, + "slice_203": 574, + "transpose_91": 570, + "mul_306": 574, + "mul_307": 571, + "linear_175": 562, + "add_241": 575, + "mul_309": 563, + "linear_176": 570, + "linear_177": 571, + "mul_310": 572, + "linear_178": 562, + "add_244": 563, + "mul_312": 562, + "linear_179": 570, + "slice_204": 574, + "slice_205": 575, + "mul_314": 571, + "transpose_92": 574, + "linear_180": 578, + "mul_316": 570, + "transpose_93": 578, + "linear_181": 571, + "transpose_94": 562, + "slice_206": 570, + "slice_207": 571, + "slice_208": 577, + "slice_209": 574, + "mul_317": 578, + "slice_210": 572, + "slice_211": 573, + "neg_28": 570, + "cat_39": 573, + "mul_318": 570, + "add_251": 572, + "mul_319": 578, + "slice_212": 569, + "slice_213": 570, + "neg_29": 577, + "cat_40": 570, + "mul_320": 577, + "add_252": 561, + "cat_41": 570, + "cat_42": 578, + "cache_update_10": 369, + "cache_update_11": 370, + "matmul_12": 574, + "transpose_96": 571, + "sigmoid_23": 574, + "mul_322": 575, + "linear_182": 561, + "add_254": 562, + "mul_324": 563, + "linear_183": 570, + "linear_184": 571, + "mul_325": 572, + "linear_185": 561, + "add_257": 563, + "mul_327": 562, + "linear_186__part_0": 579, + "linear_186__part_1": 580, + "linear_186": 581 + }, + "byteLength": 872652480 + }, + "rewrites": [ + { + "pattern": "rms_norm", + "nodes": [ + "pow_1", + "mean", + "add", + "rsqrt", + "mul_2", + "add_1", + "mul_3" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_2", + "mean_1", + "add_5", + "rsqrt_3", + "mul_9", + "mul_10" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_3", + "mean_2", + "add_7", + "rsqrt_4", + "mul_12", + "add_8", + "mul_13" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_4", + "mean_3", + "add_10", + "rsqrt_5", + "mul_15", + "add_11", + "mul_16" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_5", + "mean_4", + "add_15", + "rsqrt_8", + "mul_22", + "mul_23" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_6", + "mean_5", + "add_17", + "rsqrt_9", + "mul_25", + "add_18", + "mul_26" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_7", + "mean_6", + "add_20", + "rsqrt_10", + "mul_28", + "add_21", + "mul_29" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_8", + "mean_7", + "add_25", + "rsqrt_13", + "mul_35", + "mul_36" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_9", + "mean_8", + "add_27", + "rsqrt_14", + "mul_38", + "add_28", + "mul_39" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_10", + "mean_9", + "add_30", + "rsqrt_15", + "mul_41", + "add_31", + "mul_42" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_11", + "mean_10", + "add_32", + "rsqrt_16", + "mul_43", + "add_33", + "mul_44" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_12", + "mean_11", + "add_34", + "rsqrt_17", + "mul_45", + "add_35", + "mul_46" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_13", + "mean_12", + "add_40", + "rsqrt_18", + "mul_53", + "add_41", + "mul_54" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_14", + "mean_13", + "add_43", + "rsqrt_19", + "mul_56", + "add_44", + "mul_57" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_15", + "mean_14", + "add_48", + "rsqrt_22", + "mul_63", + "mul_64" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_16", + "mean_15", + "add_50", + "rsqrt_23", + "mul_66", + "add_51", + "mul_67" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_17", + "mean_16", + "add_53", + "rsqrt_24", + "mul_69", + "add_54", + "mul_70" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_18", + "mean_17", + "add_58", + "rsqrt_27", + "mul_76", + "mul_77" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_19", + "mean_18", + "add_60", + "rsqrt_28", + "mul_79", + "add_61", + "mul_80" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_20", + "mean_19", + "add_63", + "rsqrt_29", + "mul_82", + "add_64", + "mul_83" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_21", + "mean_20", + "add_68", + "rsqrt_32", + "mul_89", + "mul_90" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_22", + "mean_21", + "add_70", + "rsqrt_33", + "mul_92", + "add_71", + "mul_93" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_23", + "mean_22", + "add_73", + "rsqrt_34", + "mul_95", + "add_74", + "mul_96" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_24", + "mean_23", + "add_75", + "rsqrt_35", + "mul_97", + "add_76", + "mul_98" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_25", + "mean_24", + "add_77", + "rsqrt_36", + "mul_99", + "add_78", + "mul_100" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_26", + "mean_25", + "add_83", + "rsqrt_37", + "mul_107", + "add_84", + "mul_108" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_27", + "mean_26", + "add_86", + "rsqrt_38", + "mul_110", + "add_87", + "mul_111" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_28", + "mean_27", + "add_91", + "rsqrt_41", + "mul_117", + "mul_118" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_29", + "mean_28", + "add_93", + "rsqrt_42", + "mul_120", + "add_94", + "mul_121" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_30", + "mean_29", + "add_96", + "rsqrt_43", + "mul_123", + "add_97", + "mul_124" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_31", + "mean_30", + "add_101", + "rsqrt_46", + "mul_130", + "mul_131" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_32", + "mean_31", + "add_103", + "rsqrt_47", + "mul_133", + "add_104", + "mul_134" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_33", + "mean_32", + "add_106", + "rsqrt_48", + "mul_136", + "add_107", + "mul_137" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_34", + "mean_33", + "add_111", + "rsqrt_51", + "mul_143", + "mul_144" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_35", + "mean_34", + "add_113", + "rsqrt_52", + "mul_146", + "add_114", + "mul_147" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_36", + "mean_35", + "add_116", + "rsqrt_53", + "mul_149", + "add_117", + "mul_150" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_37", + "mean_36", + "add_118", + "rsqrt_54", + "mul_151", + "add_119", + "mul_152" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_38", + "mean_37", + "add_120", + "rsqrt_55", + "mul_153", + "add_121", + "mul_154" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_39", + "mean_38", + "add_126", + "rsqrt_56", + "mul_161", + "add_127", + "mul_162" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_40", + "mean_39", + "add_129", + "rsqrt_57", + "mul_164", + "add_130", + "mul_165" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_41", + "mean_40", + "add_134", + "rsqrt_60", + "mul_171", + "mul_172" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_42", + "mean_41", + "add_136", + "rsqrt_61", + "mul_174", + "add_137", + "mul_175" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_43", + "mean_42", + "add_139", + "rsqrt_62", + "mul_177", + "add_140", + "mul_178" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_44", + "mean_43", + "add_144", + "rsqrt_65", + "mul_184", + "mul_185" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_45", + "mean_44", + "add_146", + "rsqrt_66", + "mul_187", + "add_147", + "mul_188" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_46", + "mean_45", + "add_149", + "rsqrt_67", + "mul_190", + "add_150", + "mul_191" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_47", + "mean_46", + "add_154", + "rsqrt_70", + "mul_197", + "mul_198" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_48", + "mean_47", + "add_156", + "rsqrt_71", + "mul_200", + "add_157", + "mul_201" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_49", + "mean_48", + "add_159", + "rsqrt_72", + "mul_203", + "add_160", + "mul_204" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_50", + "mean_49", + "add_161", + "rsqrt_73", + "mul_205", + "add_162", + "mul_206" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_51", + "mean_50", + "add_163", + "rsqrt_74", + "mul_207", + "add_164", + "mul_208" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_52", + "mean_51", + "add_169", + "rsqrt_75", + "mul_215", + "add_170", + "mul_216" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_53", + "mean_52", + "add_172", + "rsqrt_76", + "mul_218", + "add_173", + "mul_219" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_54", + "mean_53", + "add_177", + "rsqrt_79", + "mul_225", + "mul_226" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_55", + "mean_54", + "add_179", + "rsqrt_80", + "mul_228", + "add_180", + "mul_229" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_56", + "mean_55", + "add_182", + "rsqrt_81", + "mul_231", + "add_183", + "mul_232" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_57", + "mean_56", + "add_187", + "rsqrt_84", + "mul_238", + "mul_239" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_58", + "mean_57", + "add_189", + "rsqrt_85", + "mul_241", + "add_190", + "mul_242" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_59", + "mean_58", + "add_192", + "rsqrt_86", + "mul_244", + "add_193", + "mul_245" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_60", + "mean_59", + "add_197", + "rsqrt_89", + "mul_251", + "mul_252" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_61", + "mean_60", + "add_199", + "rsqrt_90", + "mul_254", + "add_200", + "mul_255" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_62", + "mean_61", + "add_202", + "rsqrt_91", + "mul_257", + "add_203", + "mul_258" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_63", + "mean_62", + "add_204", + "rsqrt_92", + "mul_259", + "add_205", + "mul_260" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_64", + "mean_63", + "add_206", + "rsqrt_93", + "mul_261", + "add_207", + "mul_262" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_65", + "mean_64", + "add_212", + "rsqrt_94", + "mul_269", + "add_213", + "mul_270" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_66", + "mean_65", + "add_215", + "rsqrt_95", + "mul_272", + "add_216", + "mul_273" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_67", + "mean_66", + "add_220", + "rsqrt_98", + "mul_279", + "mul_280" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_68", + "mean_67", + "add_222", + "rsqrt_99", + "mul_282", + "add_223", + "mul_283" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_69", + "mean_68", + "add_225", + "rsqrt_100", + "mul_285", + "add_226", + "mul_286" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_70", + "mean_69", + "add_230", + "rsqrt_103", + "mul_292", + "mul_293" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_71", + "mean_70", + "add_232", + "rsqrt_104", + "mul_295", + "add_233", + "mul_296" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_72", + "mean_71", + "add_235", + "rsqrt_105", + "mul_298", + "add_236", + "mul_299" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_73", + "mean_72", + "add_240", + "rsqrt_108", + "mul_305", + "mul_306" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_74", + "mean_73", + "add_242", + "rsqrt_109", + "mul_308", + "add_243", + "mul_309" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_75", + "mean_74", + "add_245", + "rsqrt_110", + "mul_311", + "add_246", + "mul_312" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_76", + "mean_75", + "add_247", + "rsqrt_111", + "mul_313", + "add_248", + "mul_314" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_77", + "mean_76", + "add_249", + "rsqrt_112", + "mul_315", + "add_250", + "mul_316" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_78", + "mean_77", + "add_255", + "rsqrt_113", + "mul_323", + "add_256", + "mul_324" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_79", + "mean_78", + "add_258", + "rsqrt_114", + "mul_326", + "add_259", + "mul_327" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_1", + "mul_11" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_2", + "mul_14" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_4", + "mul_24" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_5", + "mul_27" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_7", + "mul_37" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_8", + "mul_40" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_9", + "mul_55" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_11", + "mul_65" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_12", + "mul_68" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_14", + "mul_78" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_15", + "mul_81" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_17", + "mul_91" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_18", + "mul_94" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_19", + "mul_109" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_21", + "mul_119" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_22", + "mul_122" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_24", + "mul_132" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_25", + "mul_135" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_27", + "mul_145" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_28", + "mul_148" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_29", + "mul_163" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_31", + "mul_173" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_32", + "mul_176" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_34", + "mul_186" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_35", + "mul_189" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_37", + "mul_199" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_38", + "mul_202" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_39", + "mul_217" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_41", + "mul_227" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_42", + "mul_230" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_44", + "mul_240" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_45", + "mul_243" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_47", + "mul_253" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_48", + "mul_256" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_49", + "mul_271" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_51", + "mul_281" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_52", + "mul_284" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_54", + "mul_294" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_55", + "mul_297" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_57", + "mul_307" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_58", + "mul_310" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_59", + "mul_325" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_1", + "clone_4", + "expand_2", + "unsqueeze_10", + "_unsafe_view_2", + "clone_5", + "expand_3", + "unsqueeze_11", + "transpose_25", + "matmul_1", + "mul_51", + "add_38", + "softmax", + "matmul_2" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_4", + "clone_11", + "expand_4", + "unsqueeze_17", + "_unsafe_view_5", + "clone_12", + "expand_5", + "unsqueeze_18", + "transpose_39", + "matmul_3", + "mul_105", + "add_81", + "softmax_1", + "matmul_4" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_7", + "clone_18", + "expand_6", + "unsqueeze_24", + "_unsafe_view_8", + "clone_19", + "expand_7", + "unsqueeze_25", + "transpose_53", + "matmul_5", + "mul_159", + "add_124", + "softmax_2", + "matmul_6" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_10", + "clone_25", + "expand_8", + "unsqueeze_31", + "_unsafe_view_11", + "clone_26", + "expand_9", + "unsqueeze_32", + "transpose_67", + "matmul_7", + "mul_213", + "add_167", + "softmax_3", + "matmul_8" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_13", + "clone_32", + "expand_10", + "unsqueeze_38", + "_unsafe_view_14", + "clone_33", + "expand_11", + "unsqueeze_39", + "transpose_81", + "matmul_9", + "mul_267", + "add_210", + "softmax_4", + "matmul_10" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_16", + "clone_39", + "expand_12", + "unsqueeze_45", + "_unsafe_view_17", + "clone_40", + "expand_13", + "unsqueeze_46", + "transpose_95", + "matmul_11", + "mul_321", + "add_253", + "softmax_5", + "matmul_12" + ] + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_1" + ], + "source": "b_model_model_language_model_rotary_emb_inv_freq" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_2" + ], + "source": "b_model_model_language_model_rotary_emb_inv_freq" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_3" + ], + "source": "b_model_model_language_model_rotary_emb_inv_freq" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_4" + ], + "source": "slice_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as" + ], + "source": "mul_3" + }, + { + "pattern": "elided_view", + "nodes": [ + "view" + ], + "source": "linear_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze" + ], + "source": "p_model_model_language_model_layers_0_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_5" + ], + "source": "p_model_model_language_model_layers_0_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_1" + ], + "source": "slice_13" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_2" + ], + "source": "slice_14" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_3" + ], + "source": "slice_15" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone" + ], + "source": "transpose_15" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_4" + ], + "source": "transpose_15" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_5" + ], + "source": "linear_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_6" + ], + "source": "mul_11" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_1" + ], + "source": "mul_13" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_2" + ], + "source": "mul_16" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_7" + ], + "source": "linear_9" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_1" + ], + "source": "p_model_model_language_model_layers_1_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_6" + ], + "source": "p_model_model_language_model_layers_1_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_8" + ], + "source": "slice_21" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_9" + ], + "source": "slice_22" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_10" + ], + "source": "slice_23" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_1" + ], + "source": "transpose_18" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_11" + ], + "source": "transpose_18" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_12" + ], + "source": "linear_9" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_13" + ], + "source": "mul_24" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_3" + ], + "source": "mul_26" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_4" + ], + "source": "mul_29" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_14" + ], + "source": "linear_17" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_2" + ], + "source": "p_model_model_language_model_layers_2_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_7" + ], + "source": "p_model_model_language_model_layers_2_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_15" + ], + "source": "slice_29" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_16" + ], + "source": "slice_30" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_17" + ], + "source": "slice_31" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_2" + ], + "source": "transpose_21" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_18" + ], + "source": "transpose_21" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_19" + ], + "source": "linear_17" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_20" + ], + "source": "mul_37" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_5" + ], + "source": "mul_39" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_6" + ], + "source": "mul_42" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_21" + ], + "source": "linear_24" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_3" + ], + "source": "slice_35" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view" + ], + "source": "slice_35" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_22" + ], + "source": "slice_34" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_7" + ], + "source": "mul_44" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_23" + ], + "source": "linear_25" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_8" + ], + "source": "mul_46" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_24" + ], + "source": "linear_26" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_8" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_9" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_6" + ], + "source": "transpose_26" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_25" + ], + "source": "transpose_26" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_9" + ], + "source": "mul_54" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_10" + ], + "source": "mul_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_26" + ], + "source": "linear_32" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_3" + ], + "source": "p_model_model_language_model_layers_4_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_12" + ], + "source": "p_model_model_language_model_layers_4_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_27" + ], + "source": "slice_47" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_28" + ], + "source": "slice_48" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_29" + ], + "source": "slice_49" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_7" + ], + "source": "transpose_29" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_30" + ], + "source": "transpose_29" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_31" + ], + "source": "linear_32" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_32" + ], + "source": "mul_65" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_11" + ], + "source": "mul_67" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_12" + ], + "source": "mul_70" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_33" + ], + "source": "linear_40" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_4" + ], + "source": "p_model_model_language_model_layers_5_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_13" + ], + "source": "p_model_model_language_model_layers_5_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_34" + ], + "source": "slice_55" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_35" + ], + "source": "slice_56" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_36" + ], + "source": "slice_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_8" + ], + "source": "transpose_32" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_37" + ], + "source": "transpose_32" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_38" + ], + "source": "linear_40" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_39" + ], + "source": "mul_78" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_13" + ], + "source": "mul_80" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_14" + ], + "source": "mul_83" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_40" + ], + "source": "linear_48" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_5" + ], + "source": "p_model_model_language_model_layers_6_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_14" + ], + "source": "p_model_model_language_model_layers_6_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_41" + ], + "source": "slice_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_42" + ], + "source": "slice_64" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_43" + ], + "source": "slice_65" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_9" + ], + "source": "transpose_35" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_44" + ], + "source": "transpose_35" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_45" + ], + "source": "linear_48" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_46" + ], + "source": "mul_91" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_15" + ], + "source": "mul_93" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_16" + ], + "source": "mul_96" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_47" + ], + "source": "linear_55" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_10" + ], + "source": "slice_69" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_3" + ], + "source": "slice_69" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_48" + ], + "source": "slice_68" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_17" + ], + "source": "mul_98" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_49" + ], + "source": "linear_56" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_18" + ], + "source": "mul_100" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_50" + ], + "source": "linear_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_15" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_16" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_13" + ], + "source": "transpose_40" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_51" + ], + "source": "transpose_40" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_19" + ], + "source": "mul_108" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_20" + ], + "source": "mul_111" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_52" + ], + "source": "linear_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_6" + ], + "source": "p_model_model_language_model_layers_8_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_19" + ], + "source": "p_model_model_language_model_layers_8_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_53" + ], + "source": "slice_81" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_54" + ], + "source": "slice_82" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_55" + ], + "source": "slice_83" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_14" + ], + "source": "transpose_43" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_56" + ], + "source": "transpose_43" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_57" + ], + "source": "linear_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_58" + ], + "source": "mul_119" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_21" + ], + "source": "mul_121" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_22" + ], + "source": "mul_124" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_59" + ], + "source": "linear_71" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_7" + ], + "source": "p_model_model_language_model_layers_9_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_20" + ], + "source": "p_model_model_language_model_layers_9_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_60" + ], + "source": "slice_89" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_61" + ], + "source": "slice_90" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_62" + ], + "source": "slice_91" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_15" + ], + "source": "transpose_46" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_63" + ], + "source": "transpose_46" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_64" + ], + "source": "linear_71" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_65" + ], + "source": "mul_132" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_23" + ], + "source": "mul_134" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_24" + ], + "source": "mul_137" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_66" + ], + "source": "linear_79" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_8" + ], + "source": "p_model_model_language_model_layers_10_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_21" + ], + "source": "p_model_model_language_model_layers_10_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_67" + ], + "source": "slice_97" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_68" + ], + "source": "slice_98" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_69" + ], + "source": "slice_99" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_16" + ], + "source": "transpose_49" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_70" + ], + "source": "transpose_49" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_71" + ], + "source": "linear_79" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_72" + ], + "source": "mul_145" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_25" + ], + "source": "mul_147" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_26" + ], + "source": "mul_150" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_73" + ], + "source": "linear_86" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_17" + ], + "source": "slice_103" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_6" + ], + "source": "slice_103" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_74" + ], + "source": "slice_102" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_27" + ], + "source": "mul_152" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_75" + ], + "source": "linear_87" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_28" + ], + "source": "mul_154" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_76" + ], + "source": "linear_88" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_22" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_23" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_20" + ], + "source": "transpose_54" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_77" + ], + "source": "transpose_54" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_29" + ], + "source": "mul_162" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_30" + ], + "source": "mul_165" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_78" + ], + "source": "linear_94" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_9" + ], + "source": "p_model_model_language_model_layers_12_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_26" + ], + "source": "p_model_model_language_model_layers_12_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_79" + ], + "source": "slice_115" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_80" + ], + "source": "slice_116" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_81" + ], + "source": "slice_117" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_21" + ], + "source": "transpose_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_82" + ], + "source": "transpose_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_83" + ], + "source": "linear_94" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_84" + ], + "source": "mul_173" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_31" + ], + "source": "mul_175" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_32" + ], + "source": "mul_178" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_85" + ], + "source": "linear_102" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_10" + ], + "source": "p_model_model_language_model_layers_13_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_27" + ], + "source": "p_model_model_language_model_layers_13_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_86" + ], + "source": "slice_123" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_87" + ], + "source": "slice_124" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_88" + ], + "source": "slice_125" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_22" + ], + "source": "transpose_60" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_89" + ], + "source": "transpose_60" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_90" + ], + "source": "linear_102" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_91" + ], + "source": "mul_186" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_33" + ], + "source": "mul_188" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_34" + ], + "source": "mul_191" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_92" + ], + "source": "linear_110" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_11" + ], + "source": "p_model_model_language_model_layers_14_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_28" + ], + "source": "p_model_model_language_model_layers_14_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_93" + ], + "source": "slice_131" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_94" + ], + "source": "slice_132" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_95" + ], + "source": "slice_133" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_23" + ], + "source": "transpose_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_96" + ], + "source": "transpose_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_97" + ], + "source": "linear_110" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_98" + ], + "source": "mul_199" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_35" + ], + "source": "mul_201" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_36" + ], + "source": "mul_204" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_99" + ], + "source": "linear_117" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_24" + ], + "source": "slice_137" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_9" + ], + "source": "slice_137" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_100" + ], + "source": "slice_136" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_37" + ], + "source": "mul_206" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_101" + ], + "source": "linear_118" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_38" + ], + "source": "mul_208" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_102" + ], + "source": "linear_119" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_29" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_30" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_27" + ], + "source": "transpose_68" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_103" + ], + "source": "transpose_68" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_39" + ], + "source": "mul_216" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_40" + ], + "source": "mul_219" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_104" + ], + "source": "linear_125" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_12" + ], + "source": "p_model_model_language_model_layers_16_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_33" + ], + "source": "p_model_model_language_model_layers_16_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_105" + ], + "source": "slice_149" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_106" + ], + "source": "slice_150" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_107" + ], + "source": "slice_151" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_28" + ], + "source": "transpose_71" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_108" + ], + "source": "transpose_71" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_109" + ], + "source": "linear_125" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_110" + ], + "source": "mul_227" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_41" + ], + "source": "mul_229" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_42" + ], + "source": "mul_232" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_111" + ], + "source": "linear_133" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_13" + ], + "source": "p_model_model_language_model_layers_17_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_34" + ], + "source": "p_model_model_language_model_layers_17_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_112" + ], + "source": "slice_157" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_113" + ], + "source": "slice_158" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_114" + ], + "source": "slice_159" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_29" + ], + "source": "transpose_74" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_115" + ], + "source": "transpose_74" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_116" + ], + "source": "linear_133" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_117" + ], + "source": "mul_240" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_43" + ], + "source": "mul_242" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_44" + ], + "source": "mul_245" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_118" + ], + "source": "linear_141" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_14" + ], + "source": "p_model_model_language_model_layers_18_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_35" + ], + "source": "p_model_model_language_model_layers_18_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_119" + ], + "source": "slice_165" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_120" + ], + "source": "slice_166" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_121" + ], + "source": "slice_167" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_30" + ], + "source": "transpose_77" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_122" + ], + "source": "transpose_77" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_123" + ], + "source": "linear_141" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_124" + ], + "source": "mul_253" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_45" + ], + "source": "mul_255" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_46" + ], + "source": "mul_258" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_125" + ], + "source": "linear_148" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_31" + ], + "source": "slice_171" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_12" + ], + "source": "slice_171" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_126" + ], + "source": "slice_170" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_47" + ], + "source": "mul_260" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_127" + ], + "source": "linear_149" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_48" + ], + "source": "mul_262" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_128" + ], + "source": "linear_150" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_36" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_37" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_34" + ], + "source": "transpose_82" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_129" + ], + "source": "transpose_82" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_49" + ], + "source": "mul_270" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_50" + ], + "source": "mul_273" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_130" + ], + "source": "linear_156" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_15" + ], + "source": "p_model_model_language_model_layers_20_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_40" + ], + "source": "p_model_model_language_model_layers_20_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_131" + ], + "source": "slice_183" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_132" + ], + "source": "slice_184" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_133" + ], + "source": "slice_185" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_35" + ], + "source": "transpose_85" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_134" + ], + "source": "transpose_85" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_135" + ], + "source": "linear_156" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_136" + ], + "source": "mul_281" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_51" + ], + "source": "mul_283" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_52" + ], + "source": "mul_286" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_137" + ], + "source": "linear_164" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_16" + ], + "source": "p_model_model_language_model_layers_21_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_41" + ], + "source": "p_model_model_language_model_layers_21_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_138" + ], + "source": "slice_191" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_139" + ], + "source": "slice_192" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_140" + ], + "source": "slice_193" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_36" + ], + "source": "transpose_88" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_141" + ], + "source": "transpose_88" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_142" + ], + "source": "linear_164" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_143" + ], + "source": "mul_294" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_53" + ], + "source": "mul_296" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_54" + ], + "source": "mul_299" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_144" + ], + "source": "linear_172" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_17" + ], + "source": "p_model_model_language_model_layers_22_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_42" + ], + "source": "p_model_model_language_model_layers_22_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_145" + ], + "source": "slice_199" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_146" + ], + "source": "slice_200" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_147" + ], + "source": "slice_201" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_37" + ], + "source": "transpose_91" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_148" + ], + "source": "transpose_91" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_149" + ], + "source": "linear_172" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_150" + ], + "source": "mul_307" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_55" + ], + "source": "mul_309" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_56" + ], + "source": "mul_312" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_151" + ], + "source": "linear_179" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_38" + ], + "source": "slice_205" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_15" + ], + "source": "slice_205" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_152" + ], + "source": "slice_204" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_57" + ], + "source": "mul_314" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_153" + ], + "source": "linear_180" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_58" + ], + "source": "mul_316" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_154" + ], + "source": "linear_181" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_43" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_44" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_41" + ], + "source": "transpose_96" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_155" + ], + "source": "transpose_96" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_59" + ], + "source": "mul_324" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_60" + ], + "source": "mul_327" + }, + { + "pattern": "elided_view", + "nodes": [ + "alias" + ], + "source": "mul_327" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_10" + ], + "input": "states_0", + "output": "slice_10", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_16" + ], + "input": "states_1", + "output": "slice_16", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_18" + ], + "input": "states_2", + "output": "slice_18", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_24" + ], + "input": "states_3", + "output": "slice_24", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_26" + ], + "input": "states_4", + "output": "slice_26", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_32" + ], + "input": "states_5", + "output": "slice_32", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_44" + ], + "input": "states_8", + "output": "slice_44", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_50" + ], + "input": "states_9", + "output": "slice_50", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_52" + ], + "input": "states_10", + "output": "slice_52", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_58" + ], + "input": "states_11", + "output": "slice_58", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_60" + ], + "input": "states_12", + "output": "slice_60", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_66" + ], + "input": "states_13", + "output": "slice_66", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_78" + ], + "input": "states_16", + "output": "slice_78", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_84" + ], + "input": "states_17", + "output": "slice_84", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_86" + ], + "input": "states_18", + "output": "slice_86", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_92" + ], + "input": "states_19", + "output": "slice_92", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_94" + ], + "input": "states_20", + "output": "slice_94", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_100" + ], + "input": "states_21", + "output": "slice_100", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_112" + ], + "input": "states_24", + "output": "slice_112", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_118" + ], + "input": "states_25", + "output": "slice_118", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_120" + ], + "input": "states_26", + "output": "slice_120", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_126" + ], + "input": "states_27", + "output": "slice_126", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_128" + ], + "input": "states_28", + "output": "slice_128", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_134" + ], + "input": "states_29", + "output": "slice_134", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_146" + ], + "input": "states_32", + "output": "slice_146", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_152" + ], + "input": "states_33", + "output": "slice_152", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_154" + ], + "input": "states_34", + "output": "slice_154", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_160" + ], + "input": "states_35", + "output": "slice_160", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_162" + ], + "input": "states_36", + "output": "slice_162", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_168" + ], + "input": "states_37", + "output": "slice_168", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_180" + ], + "input": "states_40", + "output": "slice_180", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_186" + ], + "input": "states_41", + "output": "slice_186", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_188" + ], + "input": "states_42", + "output": "slice_188", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_194" + ], + "input": "states_43", + "output": "slice_194", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_196" + ], + "input": "states_44", + "output": "slice_196", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_202" + ], + "input": "states_45", + "output": "slice_202", + "kind": "recurrent" + } + ], + "partitions": [ + { + "node": "embedding", + "weight": "p_model_model_language_model_embed_tokens_weight", + "rowsPerShard": 131072, + "bindingLimitBytes": 134217728 + }, + { + "node": "linear_186", + "weight": "p_model_model_language_model_embed_tokens_weight", + "rowsPerShard": 131072, + "bindingLimitBytes": 134217728 + } + ] +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/entrypoints/prefill-4.json b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/entrypoints/prefill-4.json new file mode 100644 index 0000000000000000000000000000000000000000..e5cf8d10d0513f271a93ff73ae462078c25d9dd6 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/entrypoints/prefill-4.json @@ -0,0 +1,83074 @@ +{ + "inputs": [ + "input_ids", + "position_ids", + "attention_mask", + "states_0", + "states_1", + "states_2", + "states_3", + "states_4", + "states_5", + "states_6", + "states_7", + "states_8", + "states_9", + "states_10", + "states_11", + "states_12", + "states_13", + "states_14", + "states_15", + "states_16", + "states_17", + "states_18", + "states_19", + "states_20", + "states_21", + "states_22", + "states_23", + "states_24", + "states_25", + "states_26", + "states_27", + "states_28", + "states_29", + "states_30", + "states_31", + "states_32", + "states_33", + "states_34", + "states_35", + "states_36", + "states_37", + "states_38", + "states_39", + "states_40", + "states_41", + "states_42", + "states_43", + "states_44", + "states_45", + "states_46", + "states_47" + ], + "outputs": [ + "linear_186", + "slice_10", + "slice_16", + "slice_18", + "slice_24", + "slice_26", + "slice_32", + "cache_update", + "cache_update_1", + "slice_44", + "slice_50", + "slice_52", + "slice_58", + "slice_60", + "slice_66", + "cache_update_2", + "cache_update_3", + "slice_78", + "slice_84", + "slice_86", + "slice_92", + "slice_94", + "slice_100", + "cache_update_4", + "cache_update_5", + "slice_112", + "slice_118", + "slice_120", + "slice_126", + "slice_128", + "slice_134", + "cache_update_6", + "cache_update_7", + "slice_146", + "slice_152", + "slice_154", + "slice_160", + "slice_162", + "slice_168", + "cache_update_8", + "cache_update_9", + "slice_180", + "slice_186", + "slice_188", + "slice_194", + "slice_196", + "slice_202", + "cache_update_10", + "cache_update_11" + ], + "tensors": { + "p_model_model_language_model_embed_tokens_weight": { + "shape": [ + 248320, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 254279680, + "storageDtype": "uint32", + "parts": [ + { + "path": "weights/int8-422da3921400-000.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-001.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-002.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-003.bin", + "offset": 0, + "byteLength": 52953088 + } + ], + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_embed_tokens_weight__int8_scales", + "wordsPerRow": 256 + } + }, + "p_model_model_language_model_layers_0_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 38797312, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 38801408, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 38801472, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 38801536, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 45092992, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47190144, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47206528, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47222912, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47321216, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47321728, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_0_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 49418880, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 49422976, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 51257984, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 53092992, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_1_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 54928000, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 54932096, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 54932160, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 54932224, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 61223680, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63320832, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63337216, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63353600, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63451904, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63452416, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_1_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 65549568, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 65553664, + "byteLength": 1555200 + }, + { + "path": "weights/00008.bin", + "offset": 0, + "byteLength": 279808 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 279808, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 2114816, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_2_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 3949824, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 3953920, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 3953984, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 3954048, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 10245504, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12342656, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12359040, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12375424, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12473728, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12474240, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_2_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 14571392, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 14575488, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 16410496, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 18245504, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_3_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 20080512, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 20084608, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 24278912, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 24279936, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 24804224, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 24805248, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 25329536, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_3_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 27426688, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 27430784, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 29265792, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 31100800, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_4_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 32935808, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 32939904, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 32939968, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 32940032, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 39231488, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41328640, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41345024, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41361408, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41459712, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41460224, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_4_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 43557376, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 43561472, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 45396480, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 47231488, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_5_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 49066496, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 49070592, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 49070656, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 49070720, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 55362176, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57459328, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57475712, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57492096, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57590400, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57590912, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_5_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 59688064, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 59692160, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 61527168, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 63362176, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_6_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 65197184, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 65201280, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 65201344, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 65201408, + "byteLength": 1907456 + }, + { + "path": "weights/00009.bin", + "offset": 0, + "byteLength": 4384000 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 4384000, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6481152, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6497536, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6513920, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6612224, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6612736, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_6_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 8709888, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 8713984, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 10548992, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 12384000, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_7_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 14219008, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 14223104, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 18417408, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 18418432, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 18942720, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 18943744, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 19468032, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_7_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 21565184, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 21569280, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 23404288, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 25239296, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_8_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 27074304, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 27078400, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 27078464, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 27078528, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 33369984, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35467136, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35483520, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35499904, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35598208, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35598720, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_8_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 37695872, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 37699968, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 39534976, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 41369984, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_9_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 43204992, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 43209088, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 43209152, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 43209216, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 49500672, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51597824, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51614208, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51630592, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51728896, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51729408, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_9_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 53826560, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 53830656, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 55665664, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 57500672, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_10_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 59335680, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 59339776, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 59339840, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 59339904, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 65631360, + "byteLength": 1477504 + }, + { + "path": "weights/00010.bin", + "offset": 0, + "byteLength": 619648 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 619648, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 636032, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 652416, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 750720, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 751232, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_10_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 2848384, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 2852480, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 4687488, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 6522496, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_11_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 8357504, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 8361600, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 12555904, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 12556928, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 13081216, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 13082240, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 13606528, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_11_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 15703680, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 15707776, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 17542784, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 19377792, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_12_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 21212800, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 21216896, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 21216960, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 21217024, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 27508480, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29605632, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29622016, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29638400, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29736704, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29737216, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_12_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 31834368, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 31838464, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 33673472, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 35508480, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_13_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 37343488, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 37347584, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 37347648, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 37347712, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 43639168, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45736320, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45752704, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45769088, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45867392, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45867904, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_13_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 47965056, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 47969152, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 49804160, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 51639168, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_14_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 53474176, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 53478272, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 53478336, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 53478400, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 59769856, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61867008, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61883392, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61899776, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61998080, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61998592, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_14_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 64095744, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 64099840, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 65934848, + "byteLength": 1174016 + }, + { + "path": "weights/00011.bin", + "offset": 0, + "byteLength": 660992 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 660992, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_15_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 2496000, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 2500096, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 6694400, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 6695424, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 7219712, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 7220736, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 7745024, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_15_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 9842176, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 9846272, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 11681280, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 13516288, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_16_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 15351296, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 15355392, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 15355456, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 15355520, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 21646976, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23744128, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23760512, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23776896, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23875200, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23875712, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_16_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 25972864, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 25976960, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 27811968, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 29646976, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_17_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 31481984, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 31486080, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 31486144, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 31486208, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 37777664, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 39874816, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 39891200, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 39907584, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 40005888, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 40006400, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_17_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 42103552, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 42107648, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 43942656, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 45777664, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_18_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 47612672, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 47616768, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 47616832, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 47616896, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 53908352, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56005504, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56021888, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56038272, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56136576, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56137088, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_18_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 58234240, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 58238336, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 60073344, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 61908352, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_19_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 63743360, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 63747456, + "byteLength": 3361408 + }, + { + "path": "weights/00012.bin", + "offset": 0, + "byteLength": 832896 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 832896, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 833920, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 1358208, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 1359232, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 1883520, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_19_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 3980672, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 3984768, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 5819776, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 7654784, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_20_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 9489792, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 9493888, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 9493952, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 9494016, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 15785472, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 17882624, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 17899008, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 17915392, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 18013696, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 18014208, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_20_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 20111360, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 20115456, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 21950464, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 23785472, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_21_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 25620480, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 25624576, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 25624640, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 25624704, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 31916160, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34013312, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34029696, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34046080, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34144384, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34144896, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_21_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 36242048, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 36246144, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 38081152, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 39916160, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_22_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 41751168, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 41755264, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 41755328, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 41755392, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 48046848, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50144000, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50160384, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50176768, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50275072, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50275584, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_22_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 52372736, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 52376832, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 54211840, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 56046848, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_23_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 57881856, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 57885952, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 62080256, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 62081280, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 62605568, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 62606592, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 63130880, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_23_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 65228032, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 65232128, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 67067136, + "byteLength": 41728 + }, + { + "path": "weights/00013.bin", + "offset": 0, + "byteLength": 1793280 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 1793280, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_norm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 3628288, + "byteLength": 4096 + } + ] + }, + "b_model_model_language_model_rotary_emb_inv_freq": { + "shape": [ + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 128, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 3632384, + "byteLength": 128 + } + ] + }, + "input_ids": { + "shape": [ + 1, + 4 + ], + "dtype": "int32", + "role": "input", + "byteLength": 16, + "indexBounds": [ + 0, + 248320 + ] + }, + "position_ids": { + "shape": [ + 1, + 4 + ], + "dtype": "int32", + "role": "input", + "byteLength": 16, + "indexBounds": [ + 0, + 4096 + ] + }, + "attention_mask": { + "shape": [ + 1, + 1, + 4, + 4096 + ], + "dtype": "float32", + "role": "input", + "byteLength": 65536 + }, + "states_0": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_1": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_2": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_3": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_4": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_5": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_6": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_7": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_8": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_9": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_10": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_11": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_12": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_13": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_14": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_15": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_16": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_17": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_18": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_19": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_20": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_21": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_22": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_23": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_24": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_25": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_26": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_27": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_28": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_29": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_30": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_31": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_32": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_33": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_34": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_35": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_36": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_37": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_38": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_39": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_40": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_41": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_42": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_43": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_44": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_45": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_46": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_47": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "embedding": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "unsqueeze": { + "shape": [ + 1, + 1, + 4 + ], + "dtype": "int32", + "role": "value", + "byteLength": 16 + }, + "expand": { + "shape": [ + 4, + 1, + 4 + ], + "dtype": "int32", + "role": "value", + "byteLength": 64 + }, + "slice_1": { + "shape": [ + 3, + 1, + 4 + ], + "dtype": "int32", + "role": "value", + "byteLength": 48 + }, + "unsqueeze_1": { + "shape": [ + 1, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "unsqueeze_2": { + "shape": [ + 1, + 1, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "unsqueeze_3": { + "shape": [ + 1, + 1, + 32, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "expand_1": { + "shape": [ + 3, + 1, + 32, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 384 + }, + "unsqueeze_4": { + "shape": [ + 3, + 1, + 1, + 4 + ], + "dtype": "int32", + "role": "value", + "byteLength": 48 + }, + "_to_copy": { + "shape": [ + 3, + 1, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 48 + }, + "matmul": { + "shape": [ + 3, + 1, + 32, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1536 + }, + "transpose": { + "shape": [ + 3, + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1536 + }, + "select": { + "shape": [ + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "select_1": { + "shape": [ + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "slice_2": { + "shape": [ + 1, + 4, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 176 + }, + "slice_3": { + "shape": [ + 1, + 4, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 176 + }, + "copy": { + "shape": [ + 1, + 4, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 176 + }, + "transpose_1": { + "shape": [ + 3, + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1536 + }, + "select_2": { + "shape": [ + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "slice_scatter": { + "shape": [ + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "select_scatter": { + "shape": [ + 3, + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1536 + }, + "transpose_2": { + "shape": [ + 3, + 1, + 32, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1536 + }, + "transpose_7": { + "shape": [ + 3, + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1536 + }, + "select_7": { + "shape": [ + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "slice_7": { + "shape": [ + 1, + 4, + 10 + ], + "dtype": "float32", + "role": "value", + "byteLength": 160 + }, + "transpose_8": { + "shape": [ + 3, + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1536 + }, + "select_8": { + "shape": [ + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "slice_8": { + "shape": [ + 1, + 4, + 10 + ], + "dtype": "float32", + "role": "value", + "byteLength": 160 + }, + "copy_1": { + "shape": [ + 1, + 4, + 10 + ], + "dtype": "float32", + "role": "value", + "byteLength": 160 + }, + "transpose_9": { + "shape": [ + 3, + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1536 + }, + "select_9": { + "shape": [ + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "slice_scatter_1": { + "shape": [ + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "select_scatter_1": { + "shape": [ + 3, + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1536 + }, + "transpose_10": { + "shape": [ + 3, + 1, + 32, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1536 + }, + "transpose_12": { + "shape": [ + 3, + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1536 + }, + "select_11": { + "shape": [ + 1, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "cat": { + "shape": [ + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cos": { + "shape": [ + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul": { + "shape": [ + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "sin": { + "shape": [ + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_1": { + "shape": [ + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "pow_1": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_2": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_1": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_3": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_13": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_1": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_2": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_3": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_1": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_10": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_5": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_11": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_12": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_14": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_13": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_14": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_15": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_1": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_2": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_3": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_2": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_4": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_5": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_1": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_3": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_1": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_6": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_7": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_2": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_4": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_2": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_8": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_16": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_17": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_15": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_4": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_5": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_2": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_1": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_5": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_3": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_9": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_10": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_1": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_11": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_6": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_4": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_6": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_3": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_2": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_7": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_4": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_12": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_8": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_13": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_1": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_5": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_2": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_6": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_14": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_7": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_9": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_4": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_3": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_10": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_5": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_15": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_11": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_16": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_2": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_8": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_16": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_9": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_7": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_10": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_11": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_2": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_18": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_1": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_6": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_1": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_19": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_3": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_20": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_17": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_21": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_22": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_23": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_8": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_9": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_10": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_1": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_1": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_1": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_12": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_1": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_17": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_18": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_3": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_13": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_6": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_19": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_20": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_4": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_14": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_7": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_21": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_1": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_24": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_25": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_18": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_1": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_11": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_12": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_5": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_4": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_15": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_8": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_22": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_23": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_4": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_24": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_13": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_12": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_16": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_6": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_5": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_17": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_9": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_25": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_18": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_26": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_3": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_13": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_5": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_14": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_27": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_15": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_19": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_7": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_6": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_20": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_10": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_28": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_21": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_29": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_4": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_16": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_19": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_17": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_14": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_18": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_19": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_3": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_26": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_2": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_7": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_2": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_27": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_6": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_28": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_20": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_29": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_30": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_31": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_15": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_16": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_17": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_2": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_2": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_2": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_22": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_2": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_30": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_31": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_5": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_23": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_11": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_32": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_33": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_6": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_24": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_12": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_34": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_2": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_32": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_33": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_21": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_2": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_18": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_19": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_8": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_7": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_25": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_13": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_35": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_36": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_7": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_37": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_20": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_20": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_26": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_9": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_8": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_27": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_14": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_38": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_28": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_39": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_5": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_21": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_8": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_22": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_40": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_23": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_29": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_10": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_9": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_30": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_15": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_41": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_31": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_42": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_6": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_24": { + "shape": [ + 1, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "view_21": { + "shape": [ + 1, + 4, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "slice_34": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_35": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_3": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_unsafe_view": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_22": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_11": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_10": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "add_32": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "rsqrt_16": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "mul_43": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_33": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_44": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_7": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_22": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_25": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "view_23": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "pow_12": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mean_11": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "add_34": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "rsqrt_17": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "mul_45": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_35": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_46": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "type_as_8": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "transpose_23": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "linear_26": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "view_24": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "transpose_24": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "unsqueeze_8": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "unsqueeze_9": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "slice_36": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_37": { + "shape": [ + 1, + 8, + 4, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "slice_38": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_39": { + "shape": [ + 1, + 2, + 4, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "mul_47": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_40": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_41": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "neg_3": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_4": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_48": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_36": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_49": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_42": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "slice_43": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "neg_4": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_5": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "mul_50": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "add_37": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "cat_6": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cat_7": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "cache_update": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_1": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_10": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_2": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_4": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_1": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_11": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_3": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_5": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_2": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_25": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_1": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_51": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "add_38": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "softmax": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "matmul_2": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_26": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_6": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_25": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_3": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_52": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_27": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_39": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_13": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_12": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_40": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_18": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_53": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_41": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_54": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_9": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_28": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_9": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_29": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_55": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_30": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_42": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_14": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_13": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_43": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_19": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_56": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_44": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_57": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_10": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_31": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_27": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_32": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_26": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_33": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_34": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_8": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_44": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_3": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_12": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_3": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_45": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_10": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_46": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_28": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_47": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_48": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_49": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_27": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_28": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_29": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_4": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_3": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_5": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_45": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_3": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_58": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_59": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_7": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_46": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_20": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_60": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_61": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_8": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_47": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_21": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_62": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_3": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_50": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_51": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_29": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_7": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_30": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_31": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_15": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_14": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_48": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_22": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_63": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_64": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_11": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_65": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_32": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_35": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_49": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_16": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_15": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_50": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_23": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_66": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_51": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_67": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_11": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_36": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_12": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_37": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_68": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_38": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_52": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_17": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_16": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_53": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_24": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_69": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_54": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_70": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_12": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_39": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_30": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_40": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_33": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_41": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_42": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_9": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_52": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_4": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_13": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_4": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_53": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_13": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_54": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_31": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_55": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_56": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_57": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_34": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_35": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_36": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_5": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_4": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_6": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_55": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_4": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_71": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_72": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_9": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_56": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_25": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_73": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_74": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_10": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_57": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_26": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_75": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_4": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_58": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_59": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_32": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_8": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_37": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_38": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_18": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_17": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_58": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_27": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_76": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_77": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_14": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_78": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_39": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_43": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_59": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_19": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_18": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_60": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_28": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_79": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_61": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_80": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_13": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_44": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_15": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_45": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_81": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_46": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_62": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_20": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_19": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_63": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_29": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_82": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_64": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_83": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_14": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_47": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_33": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_48": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_40": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_49": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_50": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_10": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_60": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_5": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_14": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_5": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_61": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_16": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_62": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_34": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_63": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_64": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_65": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_41": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_42": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_43": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_6": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_5": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_7": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_65": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_5": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_84": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_85": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_11": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_66": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_30": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_86": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_87": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_12": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_67": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_31": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_88": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_5": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_66": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_67": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_35": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_9": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_44": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_45": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_21": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_20": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_68": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_32": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_89": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_90": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_17": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_91": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_46": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_51": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_69": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_22": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_21": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_70": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_33": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_92": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_71": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_93": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_15": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_52": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_18": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_53": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_94": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_54": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_72": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_23": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_22": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_73": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_34": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_95": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_74": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_96": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_16": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_55": { + "shape": [ + 1, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "view_47": { + "shape": [ + 1, + 4, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "slice_68": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_69": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_10": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_unsafe_view_3": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_48": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_24": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_23": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "add_75": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "rsqrt_35": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "mul_97": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_76": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_98": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_17": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_36": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_56": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "view_49": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "pow_25": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mean_24": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "add_77": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "rsqrt_36": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "mul_99": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_78": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_100": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "type_as_18": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "transpose_37": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "linear_57": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "view_50": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "transpose_38": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "unsqueeze_15": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "unsqueeze_16": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "slice_70": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_71": { + "shape": [ + 1, + 8, + 4, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "slice_72": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_73": { + "shape": [ + 1, + 2, + 4, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "mul_101": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_74": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_75": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "neg_8": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_11": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_102": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_79": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_103": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_76": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "slice_77": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "neg_9": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_12": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "mul_104": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "add_80": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "cat_13": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cat_14": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "cache_update_2": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_3": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_17": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_4": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_11": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_4": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_18": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_5": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_12": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_5": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_39": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_3": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_105": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "add_81": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "softmax_1": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "matmul_4": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_40": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_13": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_51": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_7": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_106": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_58": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_82": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_26": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_25": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_83": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_37": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_107": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_84": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_108": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_19": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_59": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_19": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_60": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_109": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_61": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_85": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_27": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_26": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_86": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_38": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_110": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_87": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_111": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_20": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_62": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_41": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_63": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_52": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_64": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_65": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_15": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_78": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_6": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_19": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_6": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_79": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_20": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_80": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_42": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_81": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_82": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_83": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_53": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_54": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_55": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_8": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_6": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_10": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_88": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_6": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_112": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_113": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_13": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_89": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_39": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_114": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_115": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_14": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_90": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_40": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_116": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_6": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_84": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_85": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_43": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_14": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_56": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_57": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_28": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_27": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_91": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_41": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_117": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_118": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_21": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_119": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_58": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_66": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_92": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_29": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_28": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_93": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_42": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_120": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_94": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_121": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_21": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_67": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_22": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_68": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_122": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_69": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_95": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_30": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_29": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_96": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_43": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_123": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_97": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_124": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_22": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_70": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_44": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_71": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_59": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_72": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_73": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_16": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_86": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_7": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_20": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_7": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_87": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_23": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_88": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_45": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_89": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_90": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_91": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_60": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_61": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_62": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_9": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_7": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_11": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_98": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_7": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_125": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_126": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_15": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_99": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_44": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_127": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_128": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_16": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_100": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_45": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_129": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_7": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_92": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_93": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_46": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_15": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_63": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_64": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_31": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_30": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_101": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_46": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_130": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_131": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_24": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_132": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_65": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_74": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_102": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_32": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_31": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_103": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_47": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_133": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_104": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_134": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_23": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_75": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_25": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_76": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_135": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_77": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_105": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_33": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_32": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_106": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_48": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_136": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_107": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_137": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_24": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_78": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_47": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_79": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_66": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_80": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_81": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_17": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_94": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_8": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_21": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_8": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_95": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_26": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_96": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_48": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_97": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_98": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_99": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_67": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_68": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_69": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_10": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_8": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_12": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_108": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_8": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_138": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_139": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_17": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_109": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_49": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_140": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_141": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_18": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_110": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_50": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_142": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_8": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_100": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_101": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_49": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_16": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_70": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_71": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_34": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_33": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_111": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_51": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_143": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_144": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_27": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_145": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_72": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_82": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_112": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_35": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_34": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_113": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_52": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_146": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_114": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_147": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_25": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_83": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_28": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_84": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_148": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_85": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_115": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_36": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_35": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_116": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_53": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_149": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_117": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_150": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_26": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_86": { + "shape": [ + 1, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "view_73": { + "shape": [ + 1, + 4, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "slice_102": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_103": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_17": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_unsafe_view_6": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_74": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_37": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_36": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "add_118": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "rsqrt_54": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "mul_151": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_119": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_152": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_27": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_50": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_87": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "view_75": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "pow_38": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mean_37": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "add_120": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "rsqrt_55": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "mul_153": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_121": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_154": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "type_as_28": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "transpose_51": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "linear_88": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "view_76": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "transpose_52": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "unsqueeze_22": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "unsqueeze_23": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "slice_104": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_105": { + "shape": [ + 1, + 8, + 4, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "slice_106": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_107": { + "shape": [ + 1, + 2, + 4, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "mul_155": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_108": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_109": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "neg_13": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_18": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_156": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_122": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_157": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_110": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "slice_111": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "neg_14": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_19": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "mul_158": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "add_123": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "cat_20": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cat_21": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "cache_update_4": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_5": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_24": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_6": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_18": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_7": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_25": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_7": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_19": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_8": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_53": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_5": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_159": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "add_124": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "softmax_2": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "matmul_6": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_54": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_20": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_77": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_11": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_160": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_89": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_125": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_39": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_38": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_126": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_56": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_161": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_127": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_162": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_29": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_90": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_29": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_91": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_163": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_92": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_128": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_40": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_39": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_129": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_57": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_164": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_130": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_165": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_30": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_93": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_55": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_94": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_78": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_95": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_96": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_22": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_112": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_9": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_26": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_9": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_113": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_30": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_114": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_56": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_115": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_116": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_117": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_79": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_80": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_81": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_12": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_9": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_15": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_131": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_9": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_166": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_167": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_19": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_132": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_58": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_168": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_169": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_20": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_133": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_59": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_170": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_9": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_118": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_119": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_57": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_21": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_82": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_83": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_41": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_40": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_134": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_60": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_171": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_172": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_31": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_173": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_84": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_97": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_135": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_42": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_41": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_136": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_61": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_174": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_137": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_175": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_31": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_98": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_32": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_99": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_176": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_100": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_138": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_43": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_42": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_139": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_62": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_177": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_140": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_178": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_32": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_101": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_58": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_102": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_85": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_103": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_104": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_23": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_120": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_10": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_27": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_10": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_121": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_33": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_122": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_59": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_123": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_124": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_125": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_86": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_87": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_88": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_13": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_10": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_16": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_141": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_10": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_179": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_180": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_21": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_142": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_63": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_181": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_182": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_22": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_143": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_64": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_183": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_10": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_126": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_127": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_60": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_22": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_89": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_90": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_44": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_43": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_144": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_65": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_184": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_185": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_34": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_186": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_91": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_105": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_145": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_45": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_44": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_146": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_66": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_187": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_147": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_188": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_33": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_106": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_35": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_107": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_189": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_108": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_148": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_46": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_45": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_149": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_67": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_190": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_150": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_191": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_34": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_109": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_61": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_110": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_92": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_111": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_112": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_24": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_128": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_11": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_28": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_11": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_129": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_36": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_130": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_62": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_131": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_132": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_133": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_93": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_94": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_95": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_14": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_11": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_17": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_151": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_11": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_192": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_193": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_23": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_152": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_68": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_194": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_195": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_24": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_153": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_69": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_196": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_11": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_134": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_135": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_63": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_23": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_96": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_97": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_47": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_46": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_154": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_70": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_197": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_198": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_37": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_199": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_98": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_113": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_155": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_48": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_47": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_156": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_71": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_200": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_157": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_201": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_35": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_114": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_38": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_115": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_202": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_116": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_158": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_49": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_48": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_159": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_72": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_203": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_160": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_204": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_36": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_117": { + "shape": [ + 1, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "view_99": { + "shape": [ + 1, + 4, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "slice_136": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_137": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_24": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_unsafe_view_9": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_100": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_50": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_49": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "add_161": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "rsqrt_73": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "mul_205": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_162": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_206": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_37": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_64": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_118": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "view_101": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "pow_51": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mean_50": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "add_163": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "rsqrt_74": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "mul_207": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_164": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_208": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "type_as_38": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "transpose_65": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "linear_119": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "view_102": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "transpose_66": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "unsqueeze_29": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "unsqueeze_30": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "slice_138": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_139": { + "shape": [ + 1, + 8, + 4, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "slice_140": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_141": { + "shape": [ + 1, + 2, + 4, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "mul_209": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_142": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_143": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "neg_18": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_25": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_210": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_165": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_211": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_144": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "slice_145": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "neg_19": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_26": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "mul_212": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "add_166": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "cat_27": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cat_28": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "cache_update_6": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_7": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_31": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_8": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_25": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_10": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_32": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_9": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_26": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_11": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_67": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_7": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_213": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "add_167": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "softmax_3": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "matmul_8": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_68": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_27": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_103": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_15": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_214": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_120": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_168": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_52": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_51": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_169": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_75": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_215": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_170": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_216": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_39": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_121": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_39": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_122": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_217": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_123": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_171": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_53": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_52": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_172": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_76": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_218": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_173": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_219": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_40": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_124": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_69": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_125": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_104": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_126": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_127": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_29": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_146": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_12": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_33": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_12": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_147": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_40": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_148": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_70": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_149": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_150": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_151": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_105": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_106": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_107": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_16": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_12": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_20": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_174": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_12": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_220": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_221": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_25": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_175": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_77": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_222": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_223": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_26": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_176": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_78": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_224": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_12": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_152": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_153": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_71": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_28": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_108": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_109": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_54": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_53": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_177": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_79": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_225": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_226": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_41": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_227": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_110": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_128": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_178": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_55": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_54": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_179": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_80": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_228": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_180": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_229": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_41": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_129": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_42": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_130": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_230": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_131": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_181": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_56": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_55": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_182": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_81": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_231": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_183": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_232": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_42": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_132": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_72": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_133": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_111": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_134": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_135": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_30": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_154": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_13": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_34": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_13": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_155": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_43": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_156": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_73": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_157": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_158": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_159": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_112": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_113": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_114": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_17": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_13": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_21": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_184": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_13": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_233": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_234": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_27": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_185": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_82": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_235": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_236": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_28": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_186": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_83": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_237": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_13": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_160": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_161": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_74": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_29": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_115": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_116": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_57": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_56": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_187": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_84": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_238": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_239": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_44": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_240": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_117": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_136": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_188": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_58": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_57": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_189": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_85": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_241": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_190": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_242": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_43": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_137": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_45": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_138": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_243": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_139": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_191": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_59": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_58": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_192": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_86": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_244": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_193": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_245": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_44": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_140": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_75": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_141": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_118": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_142": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_143": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_31": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_162": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_14": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_35": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_14": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_163": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_46": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_164": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_76": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_165": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_166": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_167": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_119": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_120": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_121": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_18": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_14": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_22": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_194": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_14": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_246": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_247": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_29": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_195": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_87": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_248": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_249": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_30": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_196": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_88": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_250": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_14": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_168": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_169": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_77": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_30": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_122": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_123": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_60": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_59": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_197": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_89": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_251": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_252": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_47": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_253": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_124": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_144": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_198": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_61": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_60": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_199": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_90": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_254": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_200": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_255": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_45": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_145": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_48": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_146": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_256": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_147": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_201": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_62": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_61": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_202": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_91": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_257": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_203": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_258": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_46": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_148": { + "shape": [ + 1, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "view_125": { + "shape": [ + 1, + 4, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "slice_170": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_171": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_31": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_unsafe_view_12": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_126": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_63": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_62": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "add_204": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "rsqrt_92": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "mul_259": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_205": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_260": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_47": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_78": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_149": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "view_127": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "pow_64": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mean_63": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "add_206": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "rsqrt_93": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "mul_261": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_207": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_262": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "type_as_48": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "transpose_79": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "linear_150": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "view_128": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "transpose_80": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "unsqueeze_36": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "unsqueeze_37": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "slice_172": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_173": { + "shape": [ + 1, + 8, + 4, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "slice_174": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_175": { + "shape": [ + 1, + 2, + 4, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "mul_263": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_176": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_177": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "neg_23": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_32": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_264": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_208": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_265": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_178": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "slice_179": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "neg_24": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_33": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "mul_266": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "add_209": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "cat_34": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cat_35": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "cache_update_8": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_9": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_38": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_10": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_32": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_13": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_39": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_11": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_33": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_14": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_81": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_9": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_267": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "add_210": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "softmax_4": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "matmul_10": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_82": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_34": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_129": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_19": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_268": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_151": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_211": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_65": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_64": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_212": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_94": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_269": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_213": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_270": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_49": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_152": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_49": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_153": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_271": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_154": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_214": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_66": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_65": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_215": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_95": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_272": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_216": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_273": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_50": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_155": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_83": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_156": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_130": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_157": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_158": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_36": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_180": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_15": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_40": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_15": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_181": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_50": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_182": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_84": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_183": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_184": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_185": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_131": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_132": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_133": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_20": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_15": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_25": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_217": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_15": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_274": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_275": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_31": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_218": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_96": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_276": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_277": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_32": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_219": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_97": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_278": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_15": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_186": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_187": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_85": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_35": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_134": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_135": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_67": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_66": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_220": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_98": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_279": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_280": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_51": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_281": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_136": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_159": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_221": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_68": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_67": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_222": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_99": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_282": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_223": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_283": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_51": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_160": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_52": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_161": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_284": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_162": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_224": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_69": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_68": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_225": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_100": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_285": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_226": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_286": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_52": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_163": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_86": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_164": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_137": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_165": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_166": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_37": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_188": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_16": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_41": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_16": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_189": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_53": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_190": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_87": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_191": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_192": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_193": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_138": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_139": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_140": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_21": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_16": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_26": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_227": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_16": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_287": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_288": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_33": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_228": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_101": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_289": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_290": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_34": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_229": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_102": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_291": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_16": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_194": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_195": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_88": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_36": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_141": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_142": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_70": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_69": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_230": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_103": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_292": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_293": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_54": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_294": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_143": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_167": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_231": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_71": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_70": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_232": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_104": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_295": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_233": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_296": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_53": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_168": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_55": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_169": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_297": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_170": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_234": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_72": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_71": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_235": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_105": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_298": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_236": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_299": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_54": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_171": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_89": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "linear_172": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_144": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_173": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "linear_174": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "cat_38": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_196": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_17": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_42": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_17": { + "shape": [ + 1, + 6144, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 270336 + }, + "slice_197": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "silu_56": { + "shape": [ + 1, + 6144, + 8 + ], + "dtype": "float32", + "role": "value", + "byteLength": 196608 + }, + "slice_198": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "transpose_90": { + "shape": [ + 1, + 4, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "slice_199": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_200": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_201": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_145": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_146": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_147": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_22": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "exp_17": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_27": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_237": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "softplus_17": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_300": { + "shape": [ + 1, + 4, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_301": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_35": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_238": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_106": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_302": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_303": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sum_36": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_239": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_107": { + "shape": [ + 1, + 4, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_304": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "normalized_gated_delta_rule_17": { + "shape": [ + 1, + 16, + 132, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1081344 + }, + "slice_202": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_203": { + "shape": [ + 1, + 16, + 4, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_91": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_37": { + "shape": [ + 1, + 4, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_148": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_149": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_73": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_72": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_240": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_108": { + "shape": [ + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_305": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_306": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "silu_57": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_307": { + "shape": [ + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_150": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_175": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_241": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_74": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_73": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_242": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_109": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_308": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_243": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_309": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_55": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_176": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_58": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_177": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_310": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_178": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_244": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_75": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_74": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_245": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_110": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_311": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_246": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_312": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_56": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_179": { + "shape": [ + 1, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "view_151": { + "shape": [ + 1, + 4, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "slice_204": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_205": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_38": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "_unsafe_view_15": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_152": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "pow_76": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mean_75": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "add_247": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "rsqrt_111": { + "shape": [ + 1, + 4, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "mul_313": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_248": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_314": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "type_as_57": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_92": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_180": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "view_153": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "pow_77": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mean_76": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "add_249": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "rsqrt_112": { + "shape": [ + 1, + 4, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32 + }, + "mul_315": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_250": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_316": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "type_as_58": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "transpose_93": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "linear_181": { + "shape": [ + 1, + 4, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "view_154": { + "shape": [ + 1, + 4, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "transpose_94": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "unsqueeze_43": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "unsqueeze_44": { + "shape": [ + 1, + 1, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "slice_206": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_207": { + "shape": [ + 1, + 8, + 4, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "slice_208": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_209": { + "shape": [ + 1, + 2, + 4, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 6144 + }, + "mul_317": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_210": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "slice_211": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "neg_28": { + "shape": [ + 1, + 8, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_39": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_318": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "add_251": { + "shape": [ + 1, + 8, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "mul_319": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "slice_212": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "slice_213": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "neg_29": { + "shape": [ + 1, + 2, + 4, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "cat_40": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "mul_320": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "add_252": { + "shape": [ + 1, + 2, + 4, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "cat_41": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cat_42": { + "shape": [ + 1, + 2, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "cache_update_10": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_11": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_45": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_12": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_39": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_16": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_46": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_13": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_40": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_17": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_95": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_11": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_321": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "add_253": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "softmax_5": { + "shape": [ + 1, + 8, + 4, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "matmul_12": { + "shape": [ + 1, + 8, + 4, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "transpose_96": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "clone_41": { + "shape": [ + 1, + 4, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "view_155": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "sigmoid_23": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_322": { + "shape": [ + 1, + 4, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "linear_182": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_254": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_78": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_77": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_255": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_113": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_323": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_256": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_324": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_59": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_183": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "silu_59": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_184": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "mul_325": { + "shape": [ + 1, + 4, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 57344 + }, + "linear_185": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_257": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_79": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mean_78": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "add_258": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "rsqrt_114": { + "shape": [ + 1, + 4, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16 + }, + "mul_326": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "add_259": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_327": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "type_as_60": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "alias": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_186": { + "shape": [ + 1, + 4, + 248320 + ], + "dtype": "float32", + "role": "value", + "byteLength": 3973120 + }, + "p_model_model_language_model_layers_0_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 3632512, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 4091264, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 4550016, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 5008768, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 5467520, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 5926272, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 6385024, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 6843776, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 7302528, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 7761280, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 8220032, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 8678784, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 9137536, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 9596288, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 10055040, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 10513792, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 10972544, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 11431296, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 11890048, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 12348800, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 12807552, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 13266304, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 13725056, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 14183808, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 14642560, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 15101312, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 15560064, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 16018816, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 16477568, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 16936320, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 17395072, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 17853824, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 18312576, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 18771328, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 19230080, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 19688832, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 20147584, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 20606336, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 21065088, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 21523840, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 21982592, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 22441344, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 22900096, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 23358848, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 23817600, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 24276352, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 24735104, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 25193856, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 25652608, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 26111360, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 26570112, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 27028864, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 27487616, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 27946368, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 28405120, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 28863872, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 29322624, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 29781376, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 30240128, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 30698880, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 31157632, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 31616384, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 32075136, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 32533888, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 32992640, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 33451392, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 33910144, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 34368896, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 34827648, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 35286400, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 35745152, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 36203904, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 36662656, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 36664704, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 36666752, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37453184, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37715328, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37977472, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37979520, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37981568, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 38768000, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 39030144, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 39292288, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 39357824, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 39619968, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 40144256, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 40209792, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 40211840, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 40213888, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41000320, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41262464, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41524608, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41526656, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41528704, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42315136, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42577280, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42839424, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42841472, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42843520, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 43629952, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 43892096, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 44154240, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 44219776, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 44481920, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45006208, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45071744, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45073792, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45075840, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45862272, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 46124416, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 46386560, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 46388608, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 46390656, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47177088, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47439232, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47701376, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47703424, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47705472, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 48491904, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 48754048, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49016192, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49081728, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49343872, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49868160, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49933696, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49935744, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49937792, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 50724224, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 50986368, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 51248512, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 51250560, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 51252608, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52039040, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52301184, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52563328, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52565376, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52567424, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53353856, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53616000, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53878144, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53880192, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53882240, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 54668672, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 54930816, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 55192960, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 55258496, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 55520640, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56044928, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56110464, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56112512, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56114560, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56900992, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 57163136, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 57425280, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 57490816, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 57752960, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 58277248, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 58342784, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 58344832, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 58346880, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59133312, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59395456, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59657600, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59659648, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59661696, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60448128, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60710272, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60972416, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60974464, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60976512, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 61762944, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 62025088, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 62287232, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 62352768, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 62614912, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63139200, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63204736, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63206784, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63208832, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63995264, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 64257408, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 64519552, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 64521600, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 64523648, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 65310080, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 65572224, + "byteLength": 262144 + } + ] + }, + "embedding__part_0": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "embedding__part_1": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "embedding__part_2": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "embedding__part_3": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "embedding__sum_0": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "embedding__sum_1": { + "shape": [ + 1, + 4, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "linear_186__part_0": { + "shape": [ + 1, + 4, + 131072 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2097152 + }, + "linear_186__part_1": { + "shape": [ + 1, + 4, + 117248 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1875968 + }, + "linear_186__part_2": { + "shape": [ + 1, + 4, + 65536 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "linear_186__part_3": { + "shape": [ + 1, + 4, + 51712 + ], + "dtype": "float32", + "role": "value", + "byteLength": 827392 + }, + "p_model_model_language_model_embed_tokens_weight__int8_scales": { + "shape": [ + 248320, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 31784960, + "parts": [ + { + "path": "weights/int8-422da3921400-scales-000.bin", + "offset": 0, + "byteLength": 31784960 + } + ] + }, + "p_model_model_language_model_embed_tokens_weight__rows_0_131072": { + "shape": [ + 131072, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 134217728, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 0 + }, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/int8-422da3921400-000.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-001.bin", + "offset": 0, + "byteLength": 67108864 + } + ] + }, + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072": { + "shape": [ + 131072, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16777216, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight__int8_scales", + "offset": 0 + }, + "parts": [ + { + "path": "weights/int8-422da3921400-scales-000.bin", + "offset": 0, + "byteLength": 16777216 + } + ] + }, + "p_model_model_language_model_embed_tokens_weight__rows_131072_117248": { + "shape": [ + 117248, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 120061952, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 134217728 + }, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/int8-422da3921400-002.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-003.bin", + "offset": 0, + "byteLength": 52953088 + } + ] + }, + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248": { + "shape": [ + 117248, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 15007744, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight__int8_scales", + "offset": 16777216 + }, + "parts": [ + { + "path": "weights/int8-422da3921400-scales-000.bin", + "offset": 16777216, + "byteLength": 15007744 + } + ] + } + }, + "nodes": [ + { + "id": "embedding", + "op": "aten.embedding.default", + "args": [ + { + "tensor": "p_model_model_language_model_embed_tokens_weight" + }, + { + "tensor": "input_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1527, in forward\n inputs_embeds = self.get_input_embeddings()(input_ids)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "position_ids" + }, + 0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1179, in forward\n position_ids = position_ids[None, ...].expand(4, position_ids.shape[0], -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze" + }, + [ + 4, + 1, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1179, in forward\n position_ids = position_ids[None, ...].expand(4, position_ids.shape[0], -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_1", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "expand" + }, + 0, + 1, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1183, in forward\n position_ids = position_ids[1:]\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_1", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "b_model_model_language_model_rotary_emb_inv_freq" + }, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 5, in forward\n unsqueeze_1 = torch.ops.aten.unsqueeze.default(b_model_model_language_model_rotary_emb_inv_freq, 0); b_model_model_language_model_rotary_emb_inv_freq = None" + }, + { + "id": "unsqueeze_2", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "unsqueeze_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \".49\", line 6, in forward\n unsqueeze_2 = torch.ops.aten.unsqueeze.default(unsqueeze_1, 1); unsqueeze_1 = None" + }, + { + "id": "unsqueeze_3", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "unsqueeze_2" + }, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 7, in forward\n unsqueeze_3 = torch.ops.aten.unsqueeze.default(unsqueeze_2, 3); unsqueeze_2 = None" + }, + { + "id": "expand_1", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_3" + }, + [ + 3, + 1, + -1, + 1 + ] + ], + "kwargs": {}, + "source": "File \".49\", line 10, in forward\n expand_1 = torch.ops.aten.expand.default(to, [3, 1, -1, 1]); to = None" + }, + { + "id": "unsqueeze_4", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "slice_1" + }, + 2 + ], + "kwargs": {}, + "source": "File \".49\", line 13, in forward\n unsqueeze_4 = torch.ops.aten.unsqueeze.default(slice_1, 2); slice_1 = None" + }, + { + "id": "_to_copy", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "unsqueeze_4" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \".49\", line 15, in forward\n to_2 = torch.ops.aten.to.dtype(unsqueeze_4, torch.float32); unsqueeze_4 = None" + }, + { + "id": "matmul", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "expand_1" + }, + { + "tensor": "_to_copy" + } + ], + "kwargs": {}, + "source": "File \".49\", line 20, in forward\n matmul = torch.ops.aten.matmul.default(to_3, to_4); to_3 = to_4 = None" + }, + { + "id": "transpose", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 21, in forward\n transpose = torch.ops.aten.transpose.int(matmul, 2, 3); matmul = None" + }, + { + "id": "select", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 22, in forward\n select = torch.ops.aten.select.int(transpose, 0, 0)" + }, + { + "id": "select_1", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose" + }, + 0, + 1 + ], + "kwargs": {}, + "source": "File \".49\", line 23, in forward\n select_1 = torch.ops.aten.select.int(transpose, 0, 1)" + }, + { + "id": "slice_2", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "select_1" + }, + 2, + 1, + 33, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 24, in forward\n slice_2 = torch.ops.aten.slice.Tensor(select_1, 2, 1, 33, 3); select_1 = None" + }, + { + "id": "slice_3", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "select" + }, + 2, + 1, + 33, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 25, in forward\n slice_3 = torch.ops.aten.slice.Tensor(select, 2, 1, 33, 3)" + }, + { + "id": "copy", + "op": "aten.copy.default", + "args": [ + { + "tensor": "slice_3" + }, + { + "tensor": "slice_2" + } + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "transpose_1", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "select_2", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_1" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "slice_scatter", + "op": "aten.slice_scatter.default", + "args": [ + { + "tensor": "select_2" + }, + { + "tensor": "copy" + }, + 2, + 1, + 33, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "select_scatter", + "op": "aten.select_scatter.default", + "args": [ + { + "tensor": "transpose_1" + }, + { + "tensor": "slice_scatter" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "transpose_2", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "select_scatter" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "transpose_7", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "transpose_2" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 29, in forward\n slice_5 = torch.ops.aten.slice.Tensor(select, 2, 2, 30, 3)" + }, + { + "id": "select_7", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_7" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 29, in forward\n slice_5 = torch.ops.aten.slice.Tensor(select, 2, 2, 30, 3)" + }, + { + "id": "slice_7", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "select_7" + }, + 2, + 2, + 30, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 29, in forward\n slice_5 = torch.ops.aten.slice.Tensor(select, 2, 2, 30, 3)" + }, + { + "id": "transpose_8", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "transpose_2" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 28, in forward\n slice_4 = torch.ops.aten.slice.Tensor(select_2, 2, 2, 30, 3); select_2 = None" + }, + { + "id": "select_8", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_8" + }, + 0, + 2 + ], + "kwargs": {}, + "source": "File \".49\", line 28, in forward\n slice_4 = torch.ops.aten.slice.Tensor(select_2, 2, 2, 30, 3); select_2 = None" + }, + { + "id": "slice_8", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "select_8" + }, + 2, + 2, + 30, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 28, in forward\n slice_4 = torch.ops.aten.slice.Tensor(select_2, 2, 2, 30, 3); select_2 = None" + }, + { + "id": "copy_1", + "op": "aten.copy.default", + "args": [ + { + "tensor": "slice_7" + }, + { + "tensor": "slice_8" + } + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "transpose_9", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "transpose_2" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "select_9", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_9" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "slice_scatter_1", + "op": "aten.slice_scatter.default", + "args": [ + { + "tensor": "select_9" + }, + { + "tensor": "copy_1" + }, + 2, + 2, + 30, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "select_scatter_1", + "op": "aten.select_scatter.default", + "args": [ + { + "tensor": "transpose_9" + }, + { + "tensor": "slice_scatter_1" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "transpose_10", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "select_scatter_1" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "transpose_12", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "transpose_10" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 22, in forward\n select = torch.ops.aten.select.int(transpose, 0, 0)" + }, + { + "id": "select_11", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_12" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 22, in forward\n select = torch.ops.aten.select.int(transpose, 0, 0)" + }, + { + "id": "cat", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "select_11" + }, + { + "tensor": "select_11" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \".49\", line 31, in forward\n cat = torch.ops.aten.cat.default([select, select], -1); select = None" + }, + { + "id": "cos", + "op": "aten.cos.default", + "args": [ + { + "tensor": "cat" + } + ], + "kwargs": {}, + "source": "File \".49\", line 32, in forward\n cos = torch.ops.aten.cos.default(cat)" + }, + { + "id": "mul", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cos" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \".49\", line 33, in forward\n mul = torch.ops.aten.mul.Tensor(cos, 1.0); cos = None" + }, + { + "id": "sin", + "op": "aten.sin.default", + "args": [ + { + "tensor": "cat" + } + ], + "kwargs": {}, + "source": "File \".49\", line 34, in forward\n sin = torch.ops.aten.sin.default(cat); cat = None" + }, + { + "id": "mul_1", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "sin" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \".49\", line 35, in forward\n mul_1 = torch.ops.aten.mul.Tensor(sin, 1.0); sin = None" + }, + { + "id": "pow_1", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "embedding" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_1" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_2", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "embedding" + }, + { + "tensor": "rsqrt" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_1", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_3", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_2" + }, + { + "tensor": "add_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_3" + }, + { + "tensor": "embedding" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_13", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_1", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_1" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_2", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_3", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_1", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_0" + }, + { + "tensor": "transpose_13" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_10", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_1" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_5", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_1" + }, + { + "tensor": "unsqueeze_5" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_11", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_12", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_14", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_12" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_13", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_14" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_14", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_14" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_15", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_14" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_1", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_13" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_2", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_14" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_3", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_15" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_2", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_3" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_4", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg" + }, + { + "tensor": "softplus" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_5", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_1" + }, + { + "tensor": "view_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_1", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_5" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_3", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_1" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_1", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_6", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_1" + }, + { + "tensor": "rsqrt_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_7", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_2" + }, + { + "tensor": "view_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_2", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_7" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_4", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_2" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_2", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_8", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_2" + }, + { + "tensor": "rsqrt_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_6" + }, + { + "tensor": "mul_8" + }, + { + "tensor": "view_3" + }, + { + "tensor": "mul_4" + }, + { + "tensor": "sigmoid" + }, + { + "tensor": "states_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_16", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_17", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_15", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_17" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_15" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_4", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_5", + "op": "aten.view.default", + "args": [ + { + "tensor": "view" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_2", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_4" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_1", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_2" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_5", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_1" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_3", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_9", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_4" + }, + { + "tensor": "rsqrt_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_10", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_norm_weight" + }, + { + "tensor": "mul_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_1", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_11", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_10" + }, + { + "tensor": "silu_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_6", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_11" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_4", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_6" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_6", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "embedding" + }, + { + "tensor": "linear_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_3", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_6" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_2", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_3" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_7", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_2" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_4", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_12", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_6" + }, + { + "tensor": "rsqrt_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_8", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_13", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_12" + }, + { + "tensor": "add_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_1", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_13" + }, + { + "tensor": "add_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_5", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_1" + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_2", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_6", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_1" + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_14", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_2" + }, + { + "tensor": "linear_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_7", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_14" + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_9", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_6" + }, + { + "tensor": "linear_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_4", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_9" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_3", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_4" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_10", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_3" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_5", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_15", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_9" + }, + { + "tensor": "rsqrt_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_11", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_16", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_15" + }, + { + "tensor": "add_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_2", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_16" + }, + { + "tensor": "add_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_8", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_2" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_16", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_8" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_9", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_2" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_7", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_9" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_10", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_2" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_11", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_2" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_2", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_2" + }, + { + "tensor": "transpose_16" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_18", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_2" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_1", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_6", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_1", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_2" + }, + { + "tensor": "unsqueeze_6" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_19", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_1" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_3", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_20", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_3" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_17", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_20" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_21", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_17" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_22", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_17" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_23", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_17" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_8", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_21" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_9", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_22" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_10", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_23" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_1", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_1", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_1", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_12", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_11" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_1", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_17", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_1" + }, + { + "tensor": "softplus_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_18", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_8" + }, + { + "tensor": "view_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_3", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_18" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_13", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_3" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_6", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_19", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_8" + }, + { + "tensor": "rsqrt_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_20", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_9" + }, + { + "tensor": "view_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_4", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_20" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_14", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_4" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_7", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_21", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_9" + }, + { + "tensor": "rsqrt_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_1", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_19" + }, + { + "tensor": "mul_21" + }, + { + "tensor": "view_10" + }, + { + "tensor": "mul_17" + }, + { + "tensor": "sigmoid_1" + }, + { + "tensor": "states_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_24", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_1" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_25", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_1" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_18", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_25" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_1", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_18" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_11", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_1" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_12", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_7" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_5", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_11" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_4", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_5" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_15", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_4" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_8", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_22", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_11" + }, + { + "tensor": "rsqrt_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_23", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_norm_weight" + }, + { + "tensor": "mul_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_4", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_24", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_23" + }, + { + "tensor": "silu_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_13", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_24" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_12", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_13" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_16", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_9" + }, + { + "tensor": "linear_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_6", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_16" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_5", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_6" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_17", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_5" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_9", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_25", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_16" + }, + { + "tensor": "rsqrt_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_18", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_26", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_25" + }, + { + "tensor": "add_18" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_3", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_26" + }, + { + "tensor": "add_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_13", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_3" + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_5", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_14", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_3" + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_27", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_5" + }, + { + "tensor": "linear_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_15", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_27" + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_19", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_16" + }, + { + "tensor": "linear_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_7", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_19" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_6", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_7" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_20", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_6" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_10", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_20" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_28", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_19" + }, + { + "tensor": "rsqrt_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_21", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_29", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_28" + }, + { + "tensor": "add_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_4", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_29" + }, + { + "tensor": "add_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_16", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_4" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_19", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_16" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_17", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_4" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_14", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_17" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_18", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_4" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_19", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_4" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_3", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_4" + }, + { + "tensor": "transpose_19" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_26", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_3" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_2", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_7", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_2" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_2", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_3" + }, + { + "tensor": "unsqueeze_7" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_27", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_2" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_6", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_28", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_6" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_20", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_28" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_29", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_20" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_30", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_20" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_31", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_20" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_15", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_29" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_16", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_30" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_17", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_31" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_2", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_18" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_2", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_2", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_22", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_19" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_2", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_30", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_2" + }, + { + "tensor": "softplus_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_31", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_15" + }, + { + "tensor": "view_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_5", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_31" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_23", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_5" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_11", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_32", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_15" + }, + { + "tensor": "rsqrt_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_33", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_16" + }, + { + "tensor": "view_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_6", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_33" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_24", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_6" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_12", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_24" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_34", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_16" + }, + { + "tensor": "rsqrt_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_2", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_32" + }, + { + "tensor": "mul_34" + }, + { + "tensor": "view_17" + }, + { + "tensor": "mul_30" + }, + { + "tensor": "sigmoid_2" + }, + { + "tensor": "states_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_32", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_2" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_33", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_2" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_21", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_33" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_2", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_21" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_18", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_2" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_19", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_14" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_8", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_18" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_7", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_8" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_25", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_7" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_13", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_25" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_35", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_18" + }, + { + "tensor": "rsqrt_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_36", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_norm_weight" + }, + { + "tensor": "mul_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_7", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_37", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_36" + }, + { + "tensor": "silu_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_20", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_37" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_20", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_20" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_26", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_19" + }, + { + "tensor": "linear_20" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_9", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_26" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_8", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_9" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_27", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_8" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_14", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_38", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_26" + }, + { + "tensor": "rsqrt_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_28", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_39", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_38" + }, + { + "tensor": "add_28" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_5", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_39" + }, + { + "tensor": "add_26" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_21", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_5" + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_8", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_22", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_5" + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_40", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_8" + }, + { + "tensor": "linear_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_23", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_40" + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_29", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_26" + }, + { + "tensor": "linear_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_10", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_29" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_9", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_10" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_30", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_9" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_15", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_41", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_29" + }, + { + "tensor": "rsqrt_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_31", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_42", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_41" + }, + { + "tensor": "add_31" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_6", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_42" + }, + { + "tensor": "add_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_24", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_6" + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_21", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_24" + }, + [ + 1, + 4, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_34", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_21" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_35", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_21" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_3", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_35" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_3" + }, + [ + 1, + 4, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_22", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_34" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_11", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_22" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_10", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_11" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_32", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_10" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_16", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_43", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_22" + }, + { + "tensor": "rsqrt_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_33", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_44", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_43" + }, + { + "tensor": "add_33" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_7", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_44" + }, + { + "tensor": "view_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_22", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_7" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_25", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_6" + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_23", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_25" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_12", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_23" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_11", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_12" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_34", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_11" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_17", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_34" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_45", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_23" + }, + { + "tensor": "rsqrt_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_35", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_46", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_45" + }, + { + "tensor": "add_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_8", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_46" + }, + { + "tensor": "view_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_23", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_8" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_26", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_6" + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_24", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_26" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_24", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_24" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_8", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_9", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_36", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_22" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_37", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_22" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_38", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_23" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_39", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_23" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_47", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_36" + }, + { + "tensor": "unsqueeze_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_40", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_36" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_41", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_36" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_3", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_4", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_3" + }, + { + "tensor": "slice_40" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_48", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_4" + }, + { + "tensor": "unsqueeze_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_36", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_47" + }, + { + "tensor": "mul_48" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_49", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_38" + }, + { + "tensor": "unsqueeze_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_42", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_38" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_43", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_38" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_4", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_5", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_4" + }, + { + "tensor": "slice_42" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_50", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_5" + }, + { + "tensor": "unsqueeze_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_37", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_49" + }, + { + "tensor": "mul_50" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_6", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_36" + }, + { + "tensor": "slice_37" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_7", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_37" + }, + { + "tensor": "slice_39" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_6" + }, + { + "tensor": "cat_7" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_1", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_7" + }, + { + "tensor": "transpose_24" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_10", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_2", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_10" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_4", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_2" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_1", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_4" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_11", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_1" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_3", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_11" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_5", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_3" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_2", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_5" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_25", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_1" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_1", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_6" + }, + { + "tensor": "transpose_25" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_51", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_1" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_38", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_51" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_38" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_2", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax" + }, + { + "tensor": "_unsafe_view_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_26", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_2" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_6", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_26" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_25", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_6" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_3", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_52", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_25" + }, + { + "tensor": "sigmoid_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_27", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_52" + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_39", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_29" + }, + { + "tensor": "linear_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_13", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_39" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_12", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_13" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_40", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_12" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_18", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_40" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_53", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_39" + }, + { + "tensor": "rsqrt_18" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_41", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_54", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_53" + }, + { + "tensor": "add_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_9", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_54" + }, + { + "tensor": "add_39" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_28", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_9" + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_9", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_28" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_29", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_9" + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_55", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_9" + }, + { + "tensor": "linear_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_30", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_55" + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_42", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_39" + }, + { + "tensor": "linear_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_14", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_42" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_13", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_14" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_43", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_13" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_19", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_56", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_42" + }, + { + "tensor": "rsqrt_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_44", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_57", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_56" + }, + { + "tensor": "add_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_10", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_57" + }, + { + "tensor": "add_42" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_31", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_10" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_27", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_31" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_32", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_10" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_26", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_32" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_33", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_10" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_34", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_10" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_8", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_8" + }, + { + "tensor": "transpose_27" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_44", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_8" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_3", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_12", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_3" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_3", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_8" + }, + { + "tensor": "unsqueeze_12" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_45", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_3" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_10", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_46", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_10" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_28", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_46" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_47", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_28" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_48", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_28" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_49", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_28" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_27", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_47" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_28", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_48" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_29", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_49" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_4", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_33" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_3", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_5", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_45", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_34" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_3", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_58", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_5" + }, + { + "tensor": "softplus_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_59", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_27" + }, + { + "tensor": "view_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_7", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_59" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_46", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_7" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_20", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_46" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_60", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_27" + }, + { + "tensor": "rsqrt_20" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_61", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_28" + }, + { + "tensor": "view_28" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_8", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_61" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_47", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_8" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_21", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_47" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_62", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_28" + }, + { + "tensor": "rsqrt_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_3", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_60" + }, + { + "tensor": "mul_62" + }, + { + "tensor": "view_29" + }, + { + "tensor": "mul_58" + }, + { + "tensor": "sigmoid_4" + }, + { + "tensor": "states_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_50", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_3" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_51", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_3" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_29", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_51" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_7", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_29" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_30", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_7" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_31", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_26" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_15", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_30" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_14", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_15" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_48", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_14" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_22", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_48" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_63", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_30" + }, + { + "tensor": "rsqrt_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_64", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_norm_weight" + }, + { + "tensor": "mul_63" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_11", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_31" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_65", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_64" + }, + { + "tensor": "silu_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_32", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_65" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_35", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_32" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_49", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_42" + }, + { + "tensor": "linear_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_16", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_49" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_15", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_16" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_50", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_15" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_23", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_50" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_66", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_49" + }, + { + "tensor": "rsqrt_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_51", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_67", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_66" + }, + { + "tensor": "add_51" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_11", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_67" + }, + { + "tensor": "add_49" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_36", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_11" + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_12", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_36" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_37", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_11" + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_68", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_12" + }, + { + "tensor": "linear_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_38", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_68" + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_52", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_49" + }, + { + "tensor": "linear_38" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_17", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_52" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_16", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_17" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_53", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_16" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_24", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_69", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_52" + }, + { + "tensor": "rsqrt_24" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_54", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_70", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_69" + }, + { + "tensor": "add_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_12", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_70" + }, + { + "tensor": "add_52" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_39", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_12" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_30", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_39" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_40", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_12" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_33", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_40" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_41", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_12" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_42", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_12" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_9", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_10" + }, + { + "tensor": "transpose_30" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_52", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_9" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_4", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_13", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_4" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_4", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_9" + }, + { + "tensor": "unsqueeze_13" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_53", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_4" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_13", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_54", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_13" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_31", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_54" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_55", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_31" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_56", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_31" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_57", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_31" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_34", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_55" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_35", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_56" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_36", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_57" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_5", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_4", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_6", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_55", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_42" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_4", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_55" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_71", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_6" + }, + { + "tensor": "softplus_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_72", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_34" + }, + { + "tensor": "view_34" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_9", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_72" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_56", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_9" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_25", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_56" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_73", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_34" + }, + { + "tensor": "rsqrt_25" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_74", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_35" + }, + { + "tensor": "view_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_10", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_74" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_57", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_10" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_26", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_57" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_75", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_35" + }, + { + "tensor": "rsqrt_26" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_4", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_73" + }, + { + "tensor": "mul_75" + }, + { + "tensor": "view_36" + }, + { + "tensor": "mul_71" + }, + { + "tensor": "sigmoid_5" + }, + { + "tensor": "states_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_58", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_4" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_59", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_4" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_32", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_59" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_8", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_32" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_37", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_8" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_38", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_33" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_18", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_37" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_17", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_18" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_58", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_17" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_27", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_58" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_76", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_37" + }, + { + "tensor": "rsqrt_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_77", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_norm_weight" + }, + { + "tensor": "mul_76" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_14", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_38" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_78", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_77" + }, + { + "tensor": "silu_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_39", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_78" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_43", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_39" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_59", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_52" + }, + { + "tensor": "linear_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_19", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_59" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_18", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_19" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_60", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_18" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_28", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_79", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_59" + }, + { + "tensor": "rsqrt_28" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_61", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_80", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_79" + }, + { + "tensor": "add_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_13", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_80" + }, + { + "tensor": "add_59" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_44", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_13" + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_15", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_45", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_13" + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_81", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_15" + }, + { + "tensor": "linear_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_46", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_81" + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_62", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_59" + }, + { + "tensor": "linear_46" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_20", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_62" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_19", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_20" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_63", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_19" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_29", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_63" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_82", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_62" + }, + { + "tensor": "rsqrt_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_64", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_83", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_82" + }, + { + "tensor": "add_64" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_14", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_83" + }, + { + "tensor": "add_62" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_47", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_14" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_33", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_47" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_48", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_14" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_40", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_48" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_49", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_14" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_50", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_14" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_10", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_12" + }, + { + "tensor": "transpose_33" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_60", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_10" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_5", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_14", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_5" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_5", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_10" + }, + { + "tensor": "unsqueeze_14" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_61", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_5" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_16", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_62", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_16" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_34", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_62" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_63", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_34" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_64", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_34" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_65", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_34" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_41", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_63" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_42", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_64" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_43", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_65" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_6", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_49" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_5", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_7", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_65", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_50" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_5", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_65" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_84", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_7" + }, + { + "tensor": "softplus_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_85", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_41" + }, + { + "tensor": "view_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_11", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_85" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_66", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_11" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_30", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_66" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_86", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_41" + }, + { + "tensor": "rsqrt_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_87", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_42" + }, + { + "tensor": "view_42" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_12", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_87" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_67", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_12" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_31", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_88", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_42" + }, + { + "tensor": "rsqrt_31" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_5", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_86" + }, + { + "tensor": "mul_88" + }, + { + "tensor": "view_43" + }, + { + "tensor": "mul_84" + }, + { + "tensor": "sigmoid_6" + }, + { + "tensor": "states_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_66", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_5" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_67", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_5" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_35", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_67" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_9", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_35" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_44", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_9" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_45", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_40" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_21", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_44" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_20", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_21" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_68", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_20" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_32", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_68" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_89", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_44" + }, + { + "tensor": "rsqrt_32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_90", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_norm_weight" + }, + { + "tensor": "mul_89" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_17", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_91", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_90" + }, + { + "tensor": "silu_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_46", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_91" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_51", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_46" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_69", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_62" + }, + { + "tensor": "linear_51" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_22", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_69" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_21", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_22" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_70", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_21" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_33", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_70" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_92", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_69" + }, + { + "tensor": "rsqrt_33" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_71", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_93", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_92" + }, + { + "tensor": "add_71" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_15", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_93" + }, + { + "tensor": "add_69" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_52", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_15" + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_18", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_52" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_53", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_15" + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_94", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_18" + }, + { + "tensor": "linear_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_54", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_94" + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_72", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_69" + }, + { + "tensor": "linear_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_23", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_72" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_22", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_23" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_73", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_22" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_34", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_73" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_95", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_72" + }, + { + "tensor": "rsqrt_34" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_74", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_96", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_95" + }, + { + "tensor": "add_74" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_16", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_96" + }, + { + "tensor": "add_72" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_55", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_16" + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_47", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_55" + }, + [ + 1, + 4, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_68", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_47" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_69", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_47" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_10", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_69" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_3", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_10" + }, + [ + 1, + 4, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_48", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_68" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_24", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_48" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_23", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_24" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_75", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_23" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_35", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_97", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_48" + }, + { + "tensor": "rsqrt_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_76", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_98", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_97" + }, + { + "tensor": "add_76" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_17", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_98" + }, + { + "tensor": "view_48" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_36", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_17" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_56", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_16" + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_49", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_56" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_25", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_49" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_24", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_25" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_77", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_24" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_36", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_77" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_99", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_49" + }, + { + "tensor": "rsqrt_36" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_78", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_100", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_99" + }, + { + "tensor": "add_78" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_18", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_100" + }, + { + "tensor": "view_49" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_37", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_18" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_57", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_16" + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_50", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_57" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_38", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_50" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_15", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_16", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_70", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_36" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_71", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_36" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_72", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_37" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_73", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_37" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_101", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_70" + }, + { + "tensor": "unsqueeze_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_74", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_70" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_75", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_70" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_8", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_11", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_8" + }, + { + "tensor": "slice_74" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_102", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_11" + }, + { + "tensor": "unsqueeze_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_79", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_101" + }, + { + "tensor": "mul_102" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_103", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_72" + }, + { + "tensor": "unsqueeze_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_76", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_72" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_77", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_72" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_9", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_77" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_12", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_9" + }, + { + "tensor": "slice_76" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_104", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_12" + }, + { + "tensor": "unsqueeze_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_80", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_103" + }, + { + "tensor": "mul_104" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_13", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_79" + }, + { + "tensor": "slice_71" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_14", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_80" + }, + { + "tensor": "slice_73" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_2", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_14" + }, + { + "tensor": "cat_14" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_3", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_15" + }, + { + "tensor": "transpose_38" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_17", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_2" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_4", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_17" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_11", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_4" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_4", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_11" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_18", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_3" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_5", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_18" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_12", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_5" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_5", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_12" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_39", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_4" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_3", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_13" + }, + { + "tensor": "transpose_39" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_105", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_3" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_81", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_105" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_1", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_81" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_4", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_1" + }, + { + "tensor": "_unsafe_view_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_40", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_4" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_13", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_40" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_51", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_13" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_7", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_106", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_51" + }, + { + "tensor": "sigmoid_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_58", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_106" + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_82", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_72" + }, + { + "tensor": "linear_58" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_26", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_82" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_25", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_26" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_83", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_25" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_37", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_83" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_107", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_82" + }, + { + "tensor": "rsqrt_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_84", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_108", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_107" + }, + { + "tensor": "add_84" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_19", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_108" + }, + { + "tensor": "add_82" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_59", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_19" + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_19", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_59" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_60", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_19" + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_109", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_19" + }, + { + "tensor": "linear_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_61", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_109" + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_85", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_82" + }, + { + "tensor": "linear_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_27", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_85" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_26", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_27" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_86", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_26" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_38", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_86" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_110", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_85" + }, + { + "tensor": "rsqrt_38" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_87", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_111", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_110" + }, + { + "tensor": "add_87" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_20", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_111" + }, + { + "tensor": "add_85" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_62", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_20" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_41", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_62" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_63", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_20" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_52", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_63" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_64", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_20" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_65", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_20" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_15", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_16" + }, + { + "tensor": "transpose_41" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_78", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_15" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_6", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_19", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_6" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_6", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_15" + }, + { + "tensor": "unsqueeze_19" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_79", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_6" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_20", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_79" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_80", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_20" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_42", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_80" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_81", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_42" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_82", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_42" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_83", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_42" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_53", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_81" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_54", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_82" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_55", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_83" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_8", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_64" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_6", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_10", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_88", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_65" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_6", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_88" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_112", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_10" + }, + { + "tensor": "softplus_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_113", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_53" + }, + { + "tensor": "view_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_13", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_113" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_89", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_13" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_39", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_89" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_114", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_53" + }, + { + "tensor": "rsqrt_39" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_115", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_54" + }, + { + "tensor": "view_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_14", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_115" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_90", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_14" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_40", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_90" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_116", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_54" + }, + { + "tensor": "rsqrt_40" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_6", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_114" + }, + { + "tensor": "mul_116" + }, + { + "tensor": "view_55" + }, + { + "tensor": "mul_112" + }, + { + "tensor": "sigmoid_8" + }, + { + "tensor": "states_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_84", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_6" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_85", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_6" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_43", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_85" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_14", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_43" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_56", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_14" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_57", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_52" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_28", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_56" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_27", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_28" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_91", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_27" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_41", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_91" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_117", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_56" + }, + { + "tensor": "rsqrt_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_118", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_norm_weight" + }, + { + "tensor": "mul_117" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_21", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_57" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_119", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_118" + }, + { + "tensor": "silu_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_58", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_119" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_66", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_58" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_92", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_85" + }, + { + "tensor": "linear_66" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_29", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_92" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_28", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_29" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_93", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_28" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_42", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_93" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_120", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_92" + }, + { + "tensor": "rsqrt_42" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_94", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_121", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_120" + }, + { + "tensor": "add_94" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_21", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_121" + }, + { + "tensor": "add_92" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_67", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_21" + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_22", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_68", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_21" + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_122", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_22" + }, + { + "tensor": "linear_68" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_69", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_122" + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_95", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_92" + }, + { + "tensor": "linear_69" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_30", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_95" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_29", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_30" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_96", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_29" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_43", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_96" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_123", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_95" + }, + { + "tensor": "rsqrt_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_97", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_124", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_123" + }, + { + "tensor": "add_97" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_22", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_124" + }, + { + "tensor": "add_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_70", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_22" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_44", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_70" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_71", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_22" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_59", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_71" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_72", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_22" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_73", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_22" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_16", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_18" + }, + { + "tensor": "transpose_44" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_86", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_16" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_7", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_20", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_7" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_7", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_16" + }, + { + "tensor": "unsqueeze_20" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_87", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_7" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_23", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_87" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_88", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_23" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_45", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_88" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_89", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_45" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_90", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_45" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_91", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_45" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_60", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_89" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_61", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_90" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_62", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_91" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_9", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_72" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_7", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_11", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_98", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_73" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_7", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_98" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_125", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_11" + }, + { + "tensor": "softplus_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_126", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_60" + }, + { + "tensor": "view_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_15", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_126" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_99", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_15" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_44", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_99" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_127", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_60" + }, + { + "tensor": "rsqrt_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_128", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_61" + }, + { + "tensor": "view_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_16", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_128" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_100", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_16" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_45", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_100" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_129", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_61" + }, + { + "tensor": "rsqrt_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_7", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_127" + }, + { + "tensor": "mul_129" + }, + { + "tensor": "view_62" + }, + { + "tensor": "mul_125" + }, + { + "tensor": "sigmoid_9" + }, + { + "tensor": "states_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_92", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_7" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_93", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_7" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_46", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_93" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_15", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_46" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_63", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_15" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_64", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_59" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_31", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_63" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_30", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_31" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_101", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_30" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_46", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_101" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_130", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_63" + }, + { + "tensor": "rsqrt_46" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_131", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_norm_weight" + }, + { + "tensor": "mul_130" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_24", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_64" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_132", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_131" + }, + { + "tensor": "silu_24" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_65", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_132" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_74", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_65" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_102", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_95" + }, + { + "tensor": "linear_74" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_32", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_102" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_31", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_32" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_103", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_31" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_47", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_103" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_133", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_102" + }, + { + "tensor": "rsqrt_47" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_104", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_134", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_133" + }, + { + "tensor": "add_104" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_23", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_134" + }, + { + "tensor": "add_102" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_75", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_23" + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_25", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_76", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_23" + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_135", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_25" + }, + { + "tensor": "linear_76" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_77", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_135" + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_105", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_102" + }, + { + "tensor": "linear_77" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_33", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_105" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_32", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_33" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_106", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_32" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_48", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_106" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_136", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_105" + }, + { + "tensor": "rsqrt_48" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_107", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_137", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_136" + }, + { + "tensor": "add_107" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_24", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_137" + }, + { + "tensor": "add_105" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_78", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_24" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_47", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_78" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_79", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_24" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_66", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_79" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_80", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_24" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_81", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_24" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_17", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_20" + }, + { + "tensor": "transpose_47" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_94", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_17" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_8", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_21", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_8" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_8", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_17" + }, + { + "tensor": "unsqueeze_21" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_95", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_8" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_26", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_96", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_26" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_48", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_96" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_97", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_48" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_98", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_48" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_99", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_48" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_67", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_97" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_68", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_98" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_69", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_99" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_10", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_80" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_8", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_12", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_108", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_81" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_8", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_108" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_138", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_12" + }, + { + "tensor": "softplus_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_139", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_67" + }, + { + "tensor": "view_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_17", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_139" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_109", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_17" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_49", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_109" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_140", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_67" + }, + { + "tensor": "rsqrt_49" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_141", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_68" + }, + { + "tensor": "view_68" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_18", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_141" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_110", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_18" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_50", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_110" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_142", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_68" + }, + { + "tensor": "rsqrt_50" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_8", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_140" + }, + { + "tensor": "mul_142" + }, + { + "tensor": "view_69" + }, + { + "tensor": "mul_138" + }, + { + "tensor": "sigmoid_10" + }, + { + "tensor": "states_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_100", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_8" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_101", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_8" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_49", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_101" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_16", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_49" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_70", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_16" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_71", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_66" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_34", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_70" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_33", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_34" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_111", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_33" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_51", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_111" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_143", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_70" + }, + { + "tensor": "rsqrt_51" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_144", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_norm_weight" + }, + { + "tensor": "mul_143" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_27", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_71" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_145", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_144" + }, + { + "tensor": "silu_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_72", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_145" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_82", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_72" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_112", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_105" + }, + { + "tensor": "linear_82" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_35", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_112" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_34", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_35" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_113", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_34" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_52", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_146", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_112" + }, + { + "tensor": "rsqrt_52" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_114", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_147", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_146" + }, + { + "tensor": "add_114" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_25", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_147" + }, + { + "tensor": "add_112" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_83", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_25" + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_28", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_83" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_84", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_25" + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_148", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_28" + }, + { + "tensor": "linear_84" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_85", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_148" + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_115", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_112" + }, + { + "tensor": "linear_85" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_36", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_115" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_35", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_36" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_116", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_35" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_53", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_116" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_149", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_115" + }, + { + "tensor": "rsqrt_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_117", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_150", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_149" + }, + { + "tensor": "add_117" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_26", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_150" + }, + { + "tensor": "add_115" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_86", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_26" + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_73", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_86" + }, + [ + 1, + 4, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_102", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_73" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_103", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_73" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_17", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_103" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_6", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_17" + }, + [ + 1, + 4, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_74", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_102" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_37", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_74" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_36", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_37" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_118", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_36" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_54", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_118" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_151", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_74" + }, + { + "tensor": "rsqrt_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_119", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_152", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_151" + }, + { + "tensor": "add_119" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_27", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_152" + }, + { + "tensor": "view_74" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_50", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_27" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_87", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_26" + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_75", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_87" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_38", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_75" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_37", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_38" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_120", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_37" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_55", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_120" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_153", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_75" + }, + { + "tensor": "rsqrt_55" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_121", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_154", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_153" + }, + { + "tensor": "add_121" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_28", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_154" + }, + { + "tensor": "view_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_51", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_28" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_88", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_26" + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_76", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_88" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_52", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_76" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_22", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_23", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_104", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_50" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_105", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_50" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_106", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_51" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_107", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_51" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_155", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_104" + }, + { + "tensor": "unsqueeze_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_108", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_104" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_109", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_104" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_13", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_109" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_18", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_13" + }, + { + "tensor": "slice_108" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_156", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_18" + }, + { + "tensor": "unsqueeze_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_122", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_155" + }, + { + "tensor": "mul_156" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_157", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_106" + }, + { + "tensor": "unsqueeze_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_110", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_106" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_111", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_106" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_14", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_111" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_19", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_14" + }, + { + "tensor": "slice_110" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_158", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_19" + }, + { + "tensor": "unsqueeze_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_123", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_157" + }, + { + "tensor": "mul_158" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_20", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_122" + }, + { + "tensor": "slice_105" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_21", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_123" + }, + { + "tensor": "slice_107" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_4", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_22" + }, + { + "tensor": "cat_21" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_5", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_23" + }, + { + "tensor": "transpose_52" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_24", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_4" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_6", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_24" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_18", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_6" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_7", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_18" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_25", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_5" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_7", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_25" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_19", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_7" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_8", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_19" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_53", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_7" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_5", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_20" + }, + { + "tensor": "transpose_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_159", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_5" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_124", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_159" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_2", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_124" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_6", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_2" + }, + { + "tensor": "_unsafe_view_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_54", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_6" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_20", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_54" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_77", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_20" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_11", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_160", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_77" + }, + { + "tensor": "sigmoid_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_89", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_160" + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_125", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_115" + }, + { + "tensor": "linear_89" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_39", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_125" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_38", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_39" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_126", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_38" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_56", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_126" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_161", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_125" + }, + { + "tensor": "rsqrt_56" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_127", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_162", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_161" + }, + { + "tensor": "add_127" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_29", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_162" + }, + { + "tensor": "add_125" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_90", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_29" + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_29", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_90" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_91", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_29" + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_163", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_29" + }, + { + "tensor": "linear_91" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_92", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_163" + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_128", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_125" + }, + { + "tensor": "linear_92" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_40", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_128" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_39", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_40" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_129", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_39" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_57", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_129" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_164", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_128" + }, + { + "tensor": "rsqrt_57" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_130", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_165", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_164" + }, + { + "tensor": "add_130" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_30", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_165" + }, + { + "tensor": "add_128" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_93", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_30" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_55", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_93" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_94", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_30" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_78", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_94" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_95", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_30" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_96", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_30" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_22", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_24" + }, + { + "tensor": "transpose_55" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_112", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_22" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_9", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_26", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_9" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_9", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_22" + }, + { + "tensor": "unsqueeze_26" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_113", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_9" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_30", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_114", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_30" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_56", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_114" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_115", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_56" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_116", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_56" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_117", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_56" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_79", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_115" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_80", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_116" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_81", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_117" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_12", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_9", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_15", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_131", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_96" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_9", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_131" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_166", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_15" + }, + { + "tensor": "softplus_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_167", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_79" + }, + { + "tensor": "view_79" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_19", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_167" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_132", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_19" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_58", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_132" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_168", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_79" + }, + { + "tensor": "rsqrt_58" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_169", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_80" + }, + { + "tensor": "view_80" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_20", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_169" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_133", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_20" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_59", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_133" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_170", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_80" + }, + { + "tensor": "rsqrt_59" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_9", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_168" + }, + { + "tensor": "mul_170" + }, + { + "tensor": "view_81" + }, + { + "tensor": "mul_166" + }, + { + "tensor": "sigmoid_12" + }, + { + "tensor": "states_25" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_118", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_9" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_119", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_9" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_57", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_119" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_21", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_57" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_82", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_21" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_83", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_78" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_41", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_82" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_40", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_41" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_134", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_40" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_60", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_134" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_171", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_82" + }, + { + "tensor": "rsqrt_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_172", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_norm_weight" + }, + { + "tensor": "mul_171" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_31", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_83" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_173", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_172" + }, + { + "tensor": "silu_31" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_84", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_173" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_97", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_84" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_135", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_128" + }, + { + "tensor": "linear_97" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_42", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_135" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_41", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_42" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_136", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_41" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_61", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_136" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_174", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_135" + }, + { + "tensor": "rsqrt_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_137", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_175", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_174" + }, + { + "tensor": "add_137" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_31", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_175" + }, + { + "tensor": "add_135" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_98", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_31" + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_32", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_98" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_99", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_31" + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_176", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_32" + }, + { + "tensor": "linear_99" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_100", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_176" + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_138", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_135" + }, + { + "tensor": "linear_100" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_43", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_138" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_42", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_43" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_139", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_42" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_62", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_139" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_177", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_138" + }, + { + "tensor": "rsqrt_62" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_140", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_178", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_177" + }, + { + "tensor": "add_140" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_32", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_178" + }, + { + "tensor": "add_138" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_101", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_32" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_58", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_101" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_102", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_32" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_85", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_102" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_103", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_32" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_104", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_32" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_23", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_26" + }, + { + "tensor": "transpose_58" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_120", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_23" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_10", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_27", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_10" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_10", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_23" + }, + { + "tensor": "unsqueeze_27" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_121", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_10" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_33", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_121" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_122", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_33" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_59", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_122" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_123", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_59" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_124", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_59" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_125", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_59" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_86", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_123" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_87", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_124" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_88", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_125" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_13", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_103" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_10", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_16", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_141", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_104" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_10", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_141" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_179", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_16" + }, + { + "tensor": "softplus_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_180", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_86" + }, + { + "tensor": "view_86" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_21", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_180" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_142", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_21" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_63", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_142" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_181", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_86" + }, + { + "tensor": "rsqrt_63" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_182", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_87" + }, + { + "tensor": "view_87" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_22", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_182" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_143", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_22" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_64", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_143" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_183", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_87" + }, + { + "tensor": "rsqrt_64" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_10", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_181" + }, + { + "tensor": "mul_183" + }, + { + "tensor": "view_88" + }, + { + "tensor": "mul_179" + }, + { + "tensor": "sigmoid_13" + }, + { + "tensor": "states_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_126", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_10" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_127", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_10" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_60", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_127" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_22", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_60" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_89", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_22" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_90", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_85" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_44", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_89" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_43", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_44" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_144", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_43" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_65", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_144" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_184", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_89" + }, + { + "tensor": "rsqrt_65" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_185", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_norm_weight" + }, + { + "tensor": "mul_184" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_34", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_90" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_186", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_185" + }, + { + "tensor": "silu_34" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_91", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_186" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_105", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_91" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_145", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_138" + }, + { + "tensor": "linear_105" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_45", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_145" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_44", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_45" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_146", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_44" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_66", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_146" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_187", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_145" + }, + { + "tensor": "rsqrt_66" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_147", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_188", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_187" + }, + { + "tensor": "add_147" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_33", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_188" + }, + { + "tensor": "add_145" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_106", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_33" + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_35", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_106" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_107", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_33" + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_189", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_35" + }, + { + "tensor": "linear_107" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_108", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_189" + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_148", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_145" + }, + { + "tensor": "linear_108" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_46", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_148" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_45", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_46" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_149", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_45" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_67", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_149" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_190", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_148" + }, + { + "tensor": "rsqrt_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_150", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_191", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_190" + }, + { + "tensor": "add_150" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_34", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_191" + }, + { + "tensor": "add_148" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_109", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_34" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_61", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_109" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_110", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_34" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_92", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_110" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_111", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_34" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_112", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_34" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_24", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_28" + }, + { + "tensor": "transpose_61" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_128", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_24" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_11", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_28", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_11" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_11", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_24" + }, + { + "tensor": "unsqueeze_28" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_129", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_11" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_36", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_129" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_130", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_36" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_62", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_130" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_131", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_62" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_132", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_62" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_133", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_62" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_93", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_131" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_94", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_132" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_95", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_133" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_14", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_111" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_11", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_17", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_151", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_112" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_11", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_151" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_192", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_17" + }, + { + "tensor": "softplus_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_193", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_93" + }, + { + "tensor": "view_93" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_23", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_193" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_152", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_23" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_68", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_152" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_194", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_93" + }, + { + "tensor": "rsqrt_68" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_195", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_94" + }, + { + "tensor": "view_94" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_24", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_195" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_153", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_24" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_69", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_153" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_196", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_94" + }, + { + "tensor": "rsqrt_69" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_11", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_194" + }, + { + "tensor": "mul_196" + }, + { + "tensor": "view_95" + }, + { + "tensor": "mul_192" + }, + { + "tensor": "sigmoid_14" + }, + { + "tensor": "states_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_134", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_11" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_135", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_11" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_63", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_135" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_23", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_63" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_96", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_23" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_97", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_92" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_47", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_96" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_46", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_47" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_154", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_46" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_70", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_154" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_197", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_96" + }, + { + "tensor": "rsqrt_70" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_198", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_norm_weight" + }, + { + "tensor": "mul_197" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_37", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_97" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_199", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_198" + }, + { + "tensor": "silu_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_98", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_199" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_113", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_98" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_155", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_148" + }, + { + "tensor": "linear_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_48", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_155" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_47", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_48" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_156", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_47" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_71", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_156" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_200", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_155" + }, + { + "tensor": "rsqrt_71" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_157", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_201", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_200" + }, + { + "tensor": "add_157" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_35", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_201" + }, + { + "tensor": "add_155" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_114", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_35" + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_38", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_114" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_115", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_35" + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_202", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_38" + }, + { + "tensor": "linear_115" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_116", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_202" + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_158", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_155" + }, + { + "tensor": "linear_116" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_49", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_158" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_48", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_49" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_159", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_48" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_72", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_159" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_203", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_158" + }, + { + "tensor": "rsqrt_72" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_160", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_204", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_203" + }, + { + "tensor": "add_160" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_36", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_204" + }, + { + "tensor": "add_158" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_117", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_36" + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_99", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_117" + }, + [ + 1, + 4, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_136", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_99" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_137", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_99" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_24", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_137" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_9", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_24" + }, + [ + 1, + 4, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_100", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_136" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_50", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_100" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_49", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_50" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_161", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_49" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_73", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_161" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_205", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_100" + }, + { + "tensor": "rsqrt_73" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_162", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_206", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_205" + }, + { + "tensor": "add_162" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_37", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_206" + }, + { + "tensor": "view_100" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_64", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_37" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_118", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_36" + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_101", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_118" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_51", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_101" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_50", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_51" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_163", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_50" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_74", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_163" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_207", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_101" + }, + { + "tensor": "rsqrt_74" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_164", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_208", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_207" + }, + { + "tensor": "add_164" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_38", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_208" + }, + { + "tensor": "view_101" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_65", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_38" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_119", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_36" + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_102", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_119" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_66", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_102" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_29", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_30", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_138", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_64" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_139", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_64" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_140", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_65" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_141", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_65" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_209", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_138" + }, + { + "tensor": "unsqueeze_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_142", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_138" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_143", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_138" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_18", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_143" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_25", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_18" + }, + { + "tensor": "slice_142" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_210", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_25" + }, + { + "tensor": "unsqueeze_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_165", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_209" + }, + { + "tensor": "mul_210" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_211", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_140" + }, + { + "tensor": "unsqueeze_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_144", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_140" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_145", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_140" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_19", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_145" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_26", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_19" + }, + { + "tensor": "slice_144" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_212", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_26" + }, + { + "tensor": "unsqueeze_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_166", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_211" + }, + { + "tensor": "mul_212" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_27", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_165" + }, + { + "tensor": "slice_139" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_28", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_166" + }, + { + "tensor": "slice_141" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_6", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_30" + }, + { + "tensor": "cat_28" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_7", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_31" + }, + { + "tensor": "transpose_66" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_31", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_6" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_8", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_31" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_25", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_8" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_10", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_25" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_32", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_7" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_9", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_32" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_26", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_9" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_11", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_26" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_67", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_10" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_7", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_27" + }, + { + "tensor": "transpose_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_213", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_7" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_167", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_213" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_3", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_167" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_8", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_3" + }, + { + "tensor": "_unsafe_view_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_68", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_8" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_27", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_68" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_103", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_27" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_15", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_214", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_103" + }, + { + "tensor": "sigmoid_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_120", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_214" + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_168", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_158" + }, + { + "tensor": "linear_120" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_52", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_168" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_51", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_52" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_169", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_51" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_75", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_169" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_215", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_168" + }, + { + "tensor": "rsqrt_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_170", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_216", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_215" + }, + { + "tensor": "add_170" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_39", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_216" + }, + { + "tensor": "add_168" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_121", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_39" + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_39", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_121" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_122", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_39" + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_217", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_39" + }, + { + "tensor": "linear_122" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_123", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_217" + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_171", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_168" + }, + { + "tensor": "linear_123" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_53", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_171" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_52", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_53" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_172", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_52" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_76", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_172" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_218", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_171" + }, + { + "tensor": "rsqrt_76" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_173", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_219", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_218" + }, + { + "tensor": "add_173" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_40", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_219" + }, + { + "tensor": "add_171" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_124", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_40" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_69", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_124" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_125", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_40" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_104", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_125" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_126", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_40" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_127", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_40" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_29", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_32" + }, + { + "tensor": "transpose_69" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_146", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_29" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_12", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_33", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_12" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_12", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_29" + }, + { + "tensor": "unsqueeze_33" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_147", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_12" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_40", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_147" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_148", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_40" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_70", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_148" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_149", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_70" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_150", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_70" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_151", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_70" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_105", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_149" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_106", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_150" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_107", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_151" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_16", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_126" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_12", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_20", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_174", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_127" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_12", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_174" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_220", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_20" + }, + { + "tensor": "softplus_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_221", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_105" + }, + { + "tensor": "view_105" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_25", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_221" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_175", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_25" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_77", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_175" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_222", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_105" + }, + { + "tensor": "rsqrt_77" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_223", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_106" + }, + { + "tensor": "view_106" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_26", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_223" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_176", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_26" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_78", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_176" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_224", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_106" + }, + { + "tensor": "rsqrt_78" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_12", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_222" + }, + { + "tensor": "mul_224" + }, + { + "tensor": "view_107" + }, + { + "tensor": "mul_220" + }, + { + "tensor": "sigmoid_16" + }, + { + "tensor": "states_33" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_152", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_12" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_153", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_12" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_71", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_153" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_28", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_71" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_108", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_28" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_109", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_104" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_54", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_108" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_53", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_54" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_177", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_53" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_79", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_177" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_225", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_108" + }, + { + "tensor": "rsqrt_79" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_226", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_norm_weight" + }, + { + "tensor": "mul_225" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_41", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_109" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_227", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_226" + }, + { + "tensor": "silu_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_110", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_227" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_128", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_110" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_178", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_171" + }, + { + "tensor": "linear_128" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_55", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_178" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_54", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_55" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_179", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_54" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_80", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_179" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_228", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_178" + }, + { + "tensor": "rsqrt_80" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_180", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_229", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_228" + }, + { + "tensor": "add_180" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_41", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_229" + }, + { + "tensor": "add_178" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_129", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_41" + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_42", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_129" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_130", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_41" + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_230", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_42" + }, + { + "tensor": "linear_130" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_131", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_230" + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_181", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_178" + }, + { + "tensor": "linear_131" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_56", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_181" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_55", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_56" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_182", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_55" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_81", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_182" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_231", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_181" + }, + { + "tensor": "rsqrt_81" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_183", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_232", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_231" + }, + { + "tensor": "add_183" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_42", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_232" + }, + { + "tensor": "add_181" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_132", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_42" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_72", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_132" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_133", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_42" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_111", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_133" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_134", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_42" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_135", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_42" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_30", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_34" + }, + { + "tensor": "transpose_72" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_154", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_30" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_13", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_34", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_13" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_13", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_30" + }, + { + "tensor": "unsqueeze_34" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_155", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_13" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_43", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_155" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_156", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_43" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_73", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_156" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_157", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_73" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_158", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_73" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_159", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_73" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_112", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_157" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_113", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_158" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_114", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_159" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_17", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_134" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_13", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_21", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_184", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_135" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_13", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_184" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_233", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_21" + }, + { + "tensor": "softplus_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_234", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_112" + }, + { + "tensor": "view_112" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_27", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_234" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_185", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_27" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_82", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_185" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_235", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_112" + }, + { + "tensor": "rsqrt_82" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_236", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_113" + }, + { + "tensor": "view_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_28", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_236" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_186", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_28" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_83", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_186" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_237", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_113" + }, + { + "tensor": "rsqrt_83" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_13", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_235" + }, + { + "tensor": "mul_237" + }, + { + "tensor": "view_114" + }, + { + "tensor": "mul_233" + }, + { + "tensor": "sigmoid_17" + }, + { + "tensor": "states_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_160", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_13" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_161", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_13" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_74", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_161" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_29", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_74" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_115", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_29" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_116", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_111" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_57", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_115" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_56", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_57" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_187", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_56" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_84", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_187" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_238", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_115" + }, + { + "tensor": "rsqrt_84" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_239", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_norm_weight" + }, + { + "tensor": "mul_238" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_44", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_116" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_240", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_239" + }, + { + "tensor": "silu_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_117", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_240" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_136", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_117" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_188", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_181" + }, + { + "tensor": "linear_136" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_58", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_188" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_57", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_58" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_189", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_57" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_85", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_189" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_241", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_188" + }, + { + "tensor": "rsqrt_85" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_190", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_242", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_241" + }, + { + "tensor": "add_190" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_43", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_242" + }, + { + "tensor": "add_188" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_137", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_43" + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_45", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_137" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_138", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_43" + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_243", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_45" + }, + { + "tensor": "linear_138" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_139", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_243" + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_191", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_188" + }, + { + "tensor": "linear_139" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_59", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_191" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_58", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_59" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_192", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_58" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_86", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_192" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_244", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_191" + }, + { + "tensor": "rsqrt_86" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_193", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_245", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_244" + }, + { + "tensor": "add_193" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_44", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_245" + }, + { + "tensor": "add_191" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_140", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_44" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_75", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_140" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_141", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_44" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_118", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_141" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_142", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_44" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_143", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_44" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_31", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_36" + }, + { + "tensor": "transpose_75" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_162", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_31" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_14", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_35", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_14" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_14", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_31" + }, + { + "tensor": "unsqueeze_35" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_163", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_14" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_46", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_163" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_164", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_46" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_76", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_164" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_165", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_76" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_166", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_76" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_167", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_76" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_119", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_165" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_120", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_166" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_121", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_167" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_18", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_142" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_14", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_22", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_194", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_143" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_14", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_194" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_246", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_22" + }, + { + "tensor": "softplus_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_247", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_119" + }, + { + "tensor": "view_119" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_29", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_247" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_195", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_29" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_87", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_195" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_248", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_119" + }, + { + "tensor": "rsqrt_87" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_249", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_120" + }, + { + "tensor": "view_120" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_30", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_249" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_196", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_30" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_88", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_196" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_250", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_120" + }, + { + "tensor": "rsqrt_88" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_14", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_248" + }, + { + "tensor": "mul_250" + }, + { + "tensor": "view_121" + }, + { + "tensor": "mul_246" + }, + { + "tensor": "sigmoid_18" + }, + { + "tensor": "states_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_168", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_14" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_169", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_14" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_77", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_169" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_30", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_77" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_122", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_30" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_123", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_118" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_60", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_122" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_59", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_60" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_197", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_59" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_89", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_197" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_251", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_122" + }, + { + "tensor": "rsqrt_89" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_252", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_norm_weight" + }, + { + "tensor": "mul_251" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_47", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_123" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_253", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_252" + }, + { + "tensor": "silu_47" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_124", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_253" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_144", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_124" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_198", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_191" + }, + { + "tensor": "linear_144" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_61", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_198" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_60", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_61" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_199", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_60" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_90", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_199" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_254", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_198" + }, + { + "tensor": "rsqrt_90" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_200", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_255", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_254" + }, + { + "tensor": "add_200" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_45", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_255" + }, + { + "tensor": "add_198" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_145", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_45" + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_48", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_145" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_146", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_45" + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_256", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_48" + }, + { + "tensor": "linear_146" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_147", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_256" + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_201", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_198" + }, + { + "tensor": "linear_147" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_62", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_201" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_61", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_62" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_202", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_61" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_91", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_202" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_257", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_201" + }, + { + "tensor": "rsqrt_91" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_203", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_258", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_257" + }, + { + "tensor": "add_203" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_46", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_258" + }, + { + "tensor": "add_201" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_148", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_46" + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_125", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_148" + }, + [ + 1, + 4, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_170", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_125" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_171", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_125" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_31", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_171" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_12", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_31" + }, + [ + 1, + 4, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_126", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_170" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_63", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_126" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_62", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_63" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_204", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_62" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_92", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_204" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_259", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_126" + }, + { + "tensor": "rsqrt_92" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_205", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_260", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_259" + }, + { + "tensor": "add_205" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_47", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_260" + }, + { + "tensor": "view_126" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_78", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_47" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_149", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_46" + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_127", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_149" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_64", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_127" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_63", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_64" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_206", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_63" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_93", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_206" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_261", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_127" + }, + { + "tensor": "rsqrt_93" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_207", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_262", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_261" + }, + { + "tensor": "add_207" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_48", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_262" + }, + { + "tensor": "view_127" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_79", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_48" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_150", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_46" + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_128", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_150" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_80", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_128" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_36", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_37", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_172", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_78" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_173", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_78" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_174", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_79" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_175", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_79" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_263", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_172" + }, + { + "tensor": "unsqueeze_36" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_176", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_172" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_177", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_172" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_23", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_177" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_32", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_23" + }, + { + "tensor": "slice_176" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_264", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_32" + }, + { + "tensor": "unsqueeze_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_208", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_263" + }, + { + "tensor": "mul_264" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_265", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_174" + }, + { + "tensor": "unsqueeze_36" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_178", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_174" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_179", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_174" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_24", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_179" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_33", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_24" + }, + { + "tensor": "slice_178" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_266", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_33" + }, + { + "tensor": "unsqueeze_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_209", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_265" + }, + { + "tensor": "mul_266" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_34", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_208" + }, + { + "tensor": "slice_173" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_35", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_209" + }, + { + "tensor": "slice_175" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_8", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_38" + }, + { + "tensor": "cat_35" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_9", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_39" + }, + { + "tensor": "transpose_80" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_38", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_8" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_10", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_38" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_32", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_10" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_13", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_32" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_39", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_9" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_11", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_39" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_33", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_11" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_14", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_33" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_81", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_13" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_9", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_34" + }, + { + "tensor": "transpose_81" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_267", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_9" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_210", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_267" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_4", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_210" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_10", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_4" + }, + { + "tensor": "_unsafe_view_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_82", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_10" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_34", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_82" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_129", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_34" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_19", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_268", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_129" + }, + { + "tensor": "sigmoid_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_151", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_268" + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_211", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_201" + }, + { + "tensor": "linear_151" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_65", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_211" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_64", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_65" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_212", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_64" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_94", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_212" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_269", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_211" + }, + { + "tensor": "rsqrt_94" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_213", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_270", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_269" + }, + { + "tensor": "add_213" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_49", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_270" + }, + { + "tensor": "add_211" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_152", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_49" + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_49", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_152" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_153", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_49" + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_271", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_49" + }, + { + "tensor": "linear_153" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_154", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_271" + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_214", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_211" + }, + { + "tensor": "linear_154" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_66", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_214" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_65", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_66" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_215", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_65" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_95", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_215" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_272", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_214" + }, + { + "tensor": "rsqrt_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_216", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_273", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_272" + }, + { + "tensor": "add_216" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_50", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_273" + }, + { + "tensor": "add_214" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_155", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_50" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_83", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_155" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_156", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_50" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_130", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_156" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_157", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_50" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_158", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_50" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_36", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_40" + }, + { + "tensor": "transpose_83" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_180", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_36" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_15", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_40", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_15" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_15", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_36" + }, + { + "tensor": "unsqueeze_40" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_181", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_15" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_50", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_181" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_182", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_50" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_84", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_182" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_183", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_84" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_184", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_84" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_185", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_84" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_131", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_183" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_132", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_184" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_133", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_185" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_20", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_157" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_15", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_25", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_217", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_158" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_15", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_217" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_274", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_25" + }, + { + "tensor": "softplus_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_275", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_131" + }, + { + "tensor": "view_131" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_31", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_275" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_218", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_31" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_96", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_218" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_276", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_131" + }, + { + "tensor": "rsqrt_96" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_277", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_132" + }, + { + "tensor": "view_132" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_32", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_277" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_219", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_32" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_97", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_219" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_278", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_132" + }, + { + "tensor": "rsqrt_97" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_15", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_276" + }, + { + "tensor": "mul_278" + }, + { + "tensor": "view_133" + }, + { + "tensor": "mul_274" + }, + { + "tensor": "sigmoid_20" + }, + { + "tensor": "states_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_186", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_15" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_187", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_15" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_85", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_187" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_35", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_85" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_134", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_35" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_135", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_130" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_67", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_134" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_66", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_67" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_220", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_66" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_98", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_220" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_279", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_134" + }, + { + "tensor": "rsqrt_98" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_280", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_norm_weight" + }, + { + "tensor": "mul_279" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_51", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_135" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_281", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_280" + }, + { + "tensor": "silu_51" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_136", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_281" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_159", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_136" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_221", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_214" + }, + { + "tensor": "linear_159" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_68", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_221" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_67", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_68" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_222", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_67" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_99", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_222" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_282", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_221" + }, + { + "tensor": "rsqrt_99" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_223", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_283", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_282" + }, + { + "tensor": "add_223" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_51", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_283" + }, + { + "tensor": "add_221" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_160", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_51" + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_52", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_160" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_161", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_51" + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_284", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_52" + }, + { + "tensor": "linear_161" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_162", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_284" + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_224", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_221" + }, + { + "tensor": "linear_162" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_69", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_224" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_68", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_69" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_225", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_68" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_100", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_225" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_285", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_224" + }, + { + "tensor": "rsqrt_100" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_226", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_286", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_285" + }, + { + "tensor": "add_226" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_52", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_286" + }, + { + "tensor": "add_224" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_163", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_52" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_86", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_163" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_164", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_52" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_137", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_164" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_165", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_52" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_166", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_52" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_37", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_42" + }, + { + "tensor": "transpose_86" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_188", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_37" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_16", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_41", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_16" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_16", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_37" + }, + { + "tensor": "unsqueeze_41" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_189", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_16" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_53", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_189" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_190", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_53" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_87", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_190" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_191", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_87" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_192", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_87" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_193", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_87" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_138", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_191" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_139", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_192" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_140", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_193" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_21", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_165" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_16", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_26", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_227", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_166" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_16", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_227" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_287", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_26" + }, + { + "tensor": "softplus_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_288", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_138" + }, + { + "tensor": "view_138" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_33", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_288" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_228", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_33" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_101", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_228" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_289", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_138" + }, + { + "tensor": "rsqrt_101" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_290", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_139" + }, + { + "tensor": "view_139" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_34", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_290" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_229", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_34" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_102", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_229" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_291", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_139" + }, + { + "tensor": "rsqrt_102" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_16", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_289" + }, + { + "tensor": "mul_291" + }, + { + "tensor": "view_140" + }, + { + "tensor": "mul_287" + }, + { + "tensor": "sigmoid_21" + }, + { + "tensor": "states_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_194", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_16" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_195", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_16" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_88", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_195" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_36", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_88" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_141", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_36" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_142", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_137" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_70", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_141" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_69", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_70" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_230", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_69" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_103", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_230" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_292", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_141" + }, + { + "tensor": "rsqrt_103" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_293", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_norm_weight" + }, + { + "tensor": "mul_292" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_54", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_142" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_294", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_293" + }, + { + "tensor": "silu_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_143", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_294" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_167", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_143" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_231", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_224" + }, + { + "tensor": "linear_167" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_71", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_231" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_70", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_71" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_232", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_70" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_104", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_232" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_295", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_231" + }, + { + "tensor": "rsqrt_104" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_233", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_296", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_295" + }, + { + "tensor": "add_233" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_53", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_296" + }, + { + "tensor": "add_231" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_168", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_53" + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_55", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_168" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_169", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_53" + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_297", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_55" + }, + { + "tensor": "linear_169" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_170", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_297" + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_234", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_231" + }, + { + "tensor": "linear_170" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_72", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_234" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_71", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_72" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_235", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_71" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_105", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_235" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_298", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_234" + }, + { + "tensor": "rsqrt_105" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_236", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_299", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_298" + }, + { + "tensor": "add_236" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_54", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_299" + }, + { + "tensor": "add_234" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_171", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_54" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_89", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_171" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_172", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_54" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_144", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_172" + }, + [ + 1, + 4, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_173", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_54" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_174", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_54" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_38", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_44" + }, + { + "tensor": "transpose_89" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_196", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_38" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_17", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_42", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_17" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_17", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_38" + }, + { + "tensor": "unsqueeze_42" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_197", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_17" + }, + 2, + 0, + 8 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_56", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_197" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_198", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_56" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_90", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_198" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_199", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_90" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_200", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_90" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_201", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_90" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_145", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_199" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_146", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_200" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_147", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_201" + }, + [ + 1, + 4, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_22", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_173" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_17", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_27", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_237", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_174" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_17", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_237" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_300", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_27" + }, + { + "tensor": "softplus_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_301", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_145" + }, + { + "tensor": "view_145" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_35", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_301" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_238", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_35" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_106", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_238" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_302", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_145" + }, + { + "tensor": "rsqrt_106" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_303", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_146" + }, + { + "tensor": "view_146" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_36", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_303" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_239", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_36" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_107", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_239" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_304", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_146" + }, + { + "tensor": "rsqrt_107" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_17", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_302" + }, + { + "tensor": "mul_304" + }, + { + "tensor": "view_147" + }, + { + "tensor": "mul_300" + }, + { + "tensor": "sigmoid_22" + }, + { + "tensor": "states_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_202", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_17" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_203", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_17" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_91", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_203" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_37", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_91" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_148", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_37" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_149", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_144" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_73", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_148" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_72", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_73" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_240", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_72" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_108", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_240" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_305", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_148" + }, + { + "tensor": "rsqrt_108" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_306", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_norm_weight" + }, + { + "tensor": "mul_305" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_57", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_149" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_307", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_306" + }, + { + "tensor": "silu_57" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_150", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_307" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_175", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_150" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_241", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_234" + }, + { + "tensor": "linear_175" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_74", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_241" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_73", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_74" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_242", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_73" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_109", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_242" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_308", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_241" + }, + { + "tensor": "rsqrt_109" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_243", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_309", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_308" + }, + { + "tensor": "add_243" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_55", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_309" + }, + { + "tensor": "add_241" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_176", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_55" + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_58", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_176" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_177", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_55" + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_310", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_58" + }, + { + "tensor": "linear_177" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_178", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_310" + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_244", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_241" + }, + { + "tensor": "linear_178" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_75", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_244" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_74", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_75" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_245", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_74" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_110", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_245" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_311", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_244" + }, + { + "tensor": "rsqrt_110" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_246", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_312", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_311" + }, + { + "tensor": "add_246" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_56", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_312" + }, + { + "tensor": "add_244" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_179", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_56" + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_151", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_179" + }, + [ + 1, + 4, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_204", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_151" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_205", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_151" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_38", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_205" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_15", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_38" + }, + [ + 1, + 4, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_152", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_204" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_76", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_152" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_75", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_76" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_247", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_75" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_111", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_247" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_313", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_152" + }, + { + "tensor": "rsqrt_111" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_248", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_314", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_313" + }, + { + "tensor": "add_248" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_57", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_314" + }, + { + "tensor": "view_152" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_92", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_57" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_180", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_56" + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_153", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_180" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_77", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_153" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_76", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_77" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_249", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_76" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_112", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_249" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_315", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_153" + }, + { + "tensor": "rsqrt_112" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_250", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_316", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_315" + }, + { + "tensor": "add_250" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_58", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_316" + }, + { + "tensor": "view_153" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_93", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_58" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_181", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_56" + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_154", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_181" + }, + [ + 1, + 4, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_94", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_154" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_43", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_44", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_206", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_92" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_207", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_92" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_208", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_93" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_209", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_93" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_317", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_206" + }, + { + "tensor": "unsqueeze_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_210", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_206" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_211", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_206" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_28", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_211" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_39", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_28" + }, + { + "tensor": "slice_210" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_318", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_39" + }, + { + "tensor": "unsqueeze_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_251", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_317" + }, + { + "tensor": "mul_318" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_319", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_208" + }, + { + "tensor": "unsqueeze_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_212", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_208" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_213", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_208" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_29", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_213" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_40", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_29" + }, + { + "tensor": "slice_212" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_320", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_40" + }, + { + "tensor": "unsqueeze_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_252", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_319" + }, + { + "tensor": "mul_320" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_41", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_251" + }, + { + "tensor": "slice_207" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_42", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_252" + }, + { + "tensor": "slice_209" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_10", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_46" + }, + { + "tensor": "cat_42" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_11", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_47" + }, + { + "tensor": "transpose_94" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_45", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_10" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_12", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_45" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_39", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_12" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_16", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_39" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_46", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_11" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_13", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_46" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_40", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_13" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_17", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_40" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_95", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_16" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_11", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_41" + }, + { + "tensor": "transpose_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_321", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_11" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_253", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_321" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_5", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_253" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_12", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_5" + }, + { + "tensor": "_unsafe_view_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_96", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_12" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_41", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_96" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_155", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_41" + }, + [ + 1, + 4, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_23", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_322", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_155" + }, + { + "tensor": "sigmoid_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_182", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_322" + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_254", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_244" + }, + { + "tensor": "linear_182" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_78", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_254" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_77", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_78" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_255", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_77" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_113", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_255" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_323", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_254" + }, + { + "tensor": "rsqrt_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_256", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_324", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_323" + }, + { + "tensor": "add_256" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_59", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_324" + }, + { + "tensor": "add_254" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_183", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_59" + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_59", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_183" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_184", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_59" + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_325", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_59" + }, + { + "tensor": "linear_184" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_185", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_325" + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_257", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_254" + }, + { + "tensor": "linear_185" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_79", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_257" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_78", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_79" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_258", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_78" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_114", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_258" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_326", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_257" + }, + { + "tensor": "rsqrt_114" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_259", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_327", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_326" + }, + { + "tensor": "add_259" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_60", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_327" + }, + { + "tensor": "add_257" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "alias", + "op": "aten.alias.default", + "args": [ + { + "tensor": "type_as_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1787, in forward\n logits = self.lm_head(hidden_states[:, slice_indices, :])\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_186", + "op": "aten.linear.default", + "args": [ + { + "tensor": "alias" + }, + { + "tensor": "p_model_model_language_model_embed_tokens_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1787, in forward\n logits = self.lm_head(hidden_states[:, slice_indices, :])\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + } + ], + "quantization": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007083918899297714, + "meanSquaredWeightError": 4.075269544046023e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005190409719944, + "meanSquaredWeightError": 1.0724062526890066e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014917552471160889, + "meanSquaredWeightError": 8.557733188752081e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001733013428747654, + "meanSquaredWeightError": 9.211684748322568e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001013568602502346, + "meanSquaredWeightError": 4.7989616724919415e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0420619398355484, + "meanSquaredWeightError": 9.975725561162108e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03571777045726776, + "meanSquaredWeightError": 1.8712555629463168e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.015869140625, + "meanSquaredWeightError": 9.336113180324901e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006478391587734222, + "meanSquaredWeightError": 1.841971197791281e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006430316716432571, + "meanSquaredWeightError": 1.099748647703791e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009914608672261238, + "meanSquaredWeightError": 8.36655100755479e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006439946591854095, + "meanSquaredWeightError": 5.678034487033301e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010154899209737778, + "meanSquaredWeightError": 4.161362809895763e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04477539658546448, + "meanSquaredWeightError": 9.298059922002722e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02817993052303791, + "meanSquaredWeightError": 1.585794393577089e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.017333984375, + "meanSquaredWeightError": 9.087384000849852e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03647809475660324, + "meanSquaredWeightError": 9.2817742824991e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03266602009534836, + "meanSquaredWeightError": 1.564963326927682e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0213623046875, + "meanSquaredWeightError": 9.297695555687824e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00043938105227425694, + "meanSquaredWeightError": 4.2963801405448976e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0004978924989700317, + "meanSquaredWeightError": 4.300909850485368e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007458776235580444, + "meanSquaredWeightError": 6.14908657325941e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005382634699344635, + "meanSquaredWeightError": 6.678897435108411e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006161201745271683, + "meanSquaredWeightError": 2.1588792975535398e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0004498325288295746, + "meanSquaredWeightError": 7.733858886638245e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010226964950561523, + "meanSquaredWeightError": 1.0263062399928913e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007535666227340698, + "meanSquaredWeightError": 6.758249515570469e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007804818451404572, + "meanSquaredWeightError": 3.901024836494571e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02197265625, + "meanSquaredWeightError": 8.842465035741043e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03532366454601288, + "meanSquaredWeightError": 1.9045186263610958e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01679687388241291, + "meanSquaredWeightError": 8.847858339322556e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006036236882209778, + "meanSquaredWeightError": 1.5833309774393456e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00047482550144195557, + "meanSquaredWeightError": 7.293367687566388e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009035095572471619, + "meanSquaredWeightError": 9.150482149777872e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007492424920201302, + "meanSquaredWeightError": 6.225117310521e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0012094092089682817, + "meanSquaredWeightError": 4.19806900353592e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0224609375, + "meanSquaredWeightError": 9.161611842500861e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04632568359375, + "meanSquaredWeightError": 2.056171069853008e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0159912109375, + "meanSquaredWeightError": 9.076159699361597e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007881708443164825, + "meanSquaredWeightError": 1.647453373720964e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005805529654026031, + "meanSquaredWeightError": 7.469957985506426e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0016993731260299683, + "meanSquaredWeightError": 9.388108956898122e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007583741098642349, + "meanSquaredWeightError": 7.161021109425292e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006665807450190187, + "meanSquaredWeightError": 4.480360082936841e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.037353515625, + "meanSquaredWeightError": 1.04351170193695e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.05363769829273224, + "meanSquaredWeightError": 2.7039927772420924e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01563110202550888, + "meanSquaredWeightError": 9.754314760357374e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.047607421875, + "meanSquaredWeightError": 1.168099856840854e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.05196359008550644, + "meanSquaredWeightError": 2.913413027272327e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0177001953125, + "meanSquaredWeightError": 1.0753815331554506e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00047194119542837143, + "meanSquaredWeightError": 4.19777368421137e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006257314234972, + "meanSquaredWeightError": 6.4926997112024765e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000988095998764038, + "meanSquaredWeightError": 7.445749794499079e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0011749267578125, + "meanSquaredWeightError": 1.0600491151535607e-08, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005680583417415619, + "meanSquaredWeightError": 1.9959742303399253e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005613304674625397, + "meanSquaredWeightError": 9.642544540611198e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0013072118163108826, + "meanSquaredWeightError": 1.371929148064055e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000817006453871727, + "meanSquaredWeightError": 7.562183768072828e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010419189929962158, + "meanSquaredWeightError": 4.178038359725633e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0301513671875, + "meanSquaredWeightError": 1.0534593002375914e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03706229105591774, + "meanSquaredWeightError": 3.2628909139020834e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01273193396627903, + "meanSquaredWeightError": 1.0341716460970929e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008150823414325714, + "meanSquaredWeightError": 2.1580238040996846e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00043829530477523804, + "meanSquaredWeightError": 7.784422884071773e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014495849609375, + "meanSquaredWeightError": 1.5511064432871535e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000626690685749054, + "meanSquaredWeightError": 8.258425054918916e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006689839065074921, + "meanSquaredWeightError": 4.077417070647016e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0342668816447258, + "meanSquaredWeightError": 1.0280851938659907e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.046142578125, + "meanSquaredWeightError": 3.7489039641513955e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0156511589884758, + "meanSquaredWeightError": 1.021842535919859e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007631815969944, + "meanSquaredWeightError": 1.8714940708264294e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005142334848642349, + "meanSquaredWeightError": 8.547659469115843e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014417767524719238, + "meanSquaredWeightError": 1.6978646044663037e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000848725438117981, + "meanSquaredWeightError": 8.861574585239396e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009453240782022476, + "meanSquaredWeightError": 3.604574860816001e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0531267449259758, + "meanSquaredWeightError": 1.0224184734397568e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04545222967863083, + "meanSquaredWeightError": 4.110375357413432e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.014916989952325821, + "meanSquaredWeightError": 1.0372635870226077e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04470040649175644, + "meanSquaredWeightError": 1.1175421832376742e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04736328125, + "meanSquaredWeightError": 3.7854479160159826e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01680908165872097, + "meanSquaredWeightError": 1.0865893500522361e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006805174052715302, + "meanSquaredWeightError": 4.193761338200375e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008737146854400635, + "meanSquaredWeightError": 9.042077309118213e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007929764688014984, + "meanSquaredWeightError": 7.600540641305997e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014283191412687302, + "meanSquaredWeightError": 1.78832415542729e-08, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000786246731877327, + "meanSquaredWeightError": 2.793629150232846e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008093155920505524, + "meanSquaredWeightError": 1.2136558424913346e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0012167382519692183, + "meanSquaredWeightError": 8.249294580764399e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007112771272659302, + "meanSquaredWeightError": 1.0718528287156914e-08, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0012726113200187683, + "meanSquaredWeightError": 4.84854734139617e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04365234822034836, + "meanSquaredWeightError": 9.266742040381359e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.05205077677965164, + "meanSquaredWeightError": 2.190386794609367e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02226562425494194, + "meanSquaredWeightError": 9.115561283579154e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007804818451404572, + "meanSquaredWeightError": 2.823620093295176e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0004916451871395111, + "meanSquaredWeightError": 1.1404850397411792e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0022107213735580444, + "meanSquaredWeightError": 1.7722628697924847e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001212533563375473, + "meanSquaredWeightError": 9.309404802593235e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001015007495880127, + "meanSquaredWeightError": 3.86098308879923e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03144530951976776, + "meanSquaredWeightError": 1.1349436590535333e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.043701171875, + "meanSquaredWeightError": 3.7038337268313626e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01273193396627903, + "meanSquaredWeightError": 1.1466236173873767e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000811239704489708, + "meanSquaredWeightError": 2.8647622940525252e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0004772273823618889, + "meanSquaredWeightError": 9.231112763075089e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001953125, + "meanSquaredWeightError": 1.5947463793963834e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010938271880149841, + "meanSquaredWeightError": 9.523899890950815e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010099634528160095, + "meanSquaredWeightError": 4.4895744899520196e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02999267354607582, + "meanSquaredWeightError": 1.2448520010366337e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03925780951976776, + "meanSquaredWeightError": 3.853431735478807e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02015380933880806, + "meanSquaredWeightError": 1.2184170827822527e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008477643132209778, + "meanSquaredWeightError": 3.4146133032209036e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007651001214981079, + "meanSquaredWeightError": 1.287605755351251e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0015244372189044952, + "meanSquaredWeightError": 1.2338504440378983e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001476377248764038, + "meanSquaredWeightError": 9.998956329582143e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001247616484761238, + "meanSquaredWeightError": 5.051828733115826e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0316162109375, + "meanSquaredWeightError": 1.3124330280334107e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.05973249301314354, + "meanSquaredWeightError": 4.17917181039229e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02794799953699112, + "meanSquaredWeightError": 1.3120700259605655e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0629359632730484, + "meanSquaredWeightError": 1.335452111561608e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0677926167845726, + "meanSquaredWeightError": 4.988436103303684e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0322265625, + "meanSquaredWeightError": 1.6804390270408476e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000582478940486908, + "meanSquaredWeightError": 3.9153045250373e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010428838431835175, + "meanSquaredWeightError": 8.011205920865905e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001420147716999054, + "meanSquaredWeightError": 8.69789218427286e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.002052607014775276, + "meanSquaredWeightError": 1.3341496796215324e-08, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007699094712734222, + "meanSquaredWeightError": 2.5097170563981308e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006843628361821175, + "meanSquaredWeightError": 1.136356786446413e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009746402502059937, + "meanSquaredWeightError": 8.813485941061572e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000755971297621727, + "meanSquaredWeightError": 1.1121874976538493e-08, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0011303499341011047, + "meanSquaredWeightError": 4.848550450020639e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03278808295726776, + "meanSquaredWeightError": 8.690146273693244e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0361328125, + "meanSquaredWeightError": 3.182140517310472e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01421508751809597, + "meanSquaredWeightError": 8.130265882755339e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03572387620806694, + "meanSquaredWeightError": 8.140267482303898e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02766112983226776, + "meanSquaredWeightError": 3.193840257154079e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01599992997944355, + "meanSquaredWeightError": 7.776296229167201e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00040658097714185715, + "meanSquaredWeightError": 4.779415085920391e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0012706853449344635, + "meanSquaredWeightError": 4.892196425743123e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009188912808895111, + "meanSquaredWeightError": 1.1496103624608622e-08, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005174772813916206, + "meanSquaredWeightError": 5.1614277296607725e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006728284060955048, + "meanSquaredWeightError": 2.26705694217344e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005107489414513111, + "meanSquaredWeightError": 8.198658640878875e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001364879310131073, + "meanSquaredWeightError": 1.0674462203041912e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006516836583614349, + "meanSquaredWeightError": 9.449681925843834e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009464053437113762, + "meanSquaredWeightError": 4.404944853320103e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0265415720641613, + "meanSquaredWeightError": 8.238882287514571e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0333251990377903, + "meanSquaredWeightError": 2.927112973338808e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0194091796875, + "meanSquaredWeightError": 8.087501441877976e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005795937031507492, + "meanSquaredWeightError": 1.80652719450336e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005795937031507492, + "meanSquaredWeightError": 7.52482787191866e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0013129785656929016, + "meanSquaredWeightError": 9.38157196372913e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008208509534597397, + "meanSquaredWeightError": 8.128849593447285e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010572969913482666, + "meanSquaredWeightError": 4.412351817251192e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04150390625, + "meanSquaredWeightError": 8.620316975793685e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0302734375, + "meanSquaredWeightError": 2.6132290713576367e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02303466945886612, + "meanSquaredWeightError": 8.573984473514429e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006062665488570929, + "meanSquaredWeightError": 1.623409140449894e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005637351423501968, + "meanSquaredWeightError": 9.469766304448513e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008612200617790222, + "meanSquaredWeightError": 8.620569147410606e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008112378418445587, + "meanSquaredWeightError": 7.537758861531074e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008592978119850159, + "meanSquaredWeightError": 4.527719088542881e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.028881840407848358, + "meanSquaredWeightError": 9.06856655547017e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04156494140625, + "meanSquaredWeightError": 2.46402237280563e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01636962778866291, + "meanSquaredWeightError": 9.115400416703778e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02463379129767418, + "meanSquaredWeightError": 8.773882314017101e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.030029296875, + "meanSquaredWeightError": 1.9424107904342236e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02099609375, + "meanSquaredWeightError": 9.083250915864483e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00036885414738208055, + "meanSquaredWeightError": 3.3473028793196136e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005767084658145905, + "meanSquaredWeightError": 3.908651180495326e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001015007495880127, + "meanSquaredWeightError": 7.261003798220145e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005060629919171333, + "meanSquaredWeightError": 4.816435694721122e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005863220430910587, + "meanSquaredWeightError": 1.6181449069563314e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00045013427734375, + "meanSquaredWeightError": 6.057955026506079e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009198524057865143, + "meanSquaredWeightError": 8.711624310819843e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0013682451099157333, + "meanSquaredWeightError": 6.523535933666835e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008987076580524445, + "meanSquaredWeightError": 4.059202307615806e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.037841796875, + "meanSquaredWeightError": 8.265481596936297e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03264160454273224, + "meanSquaredWeightError": 1.6808718328320538e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0196533203125, + "meanSquaredWeightError": 8.899761496650171e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000939076766371727, + "meanSquaredWeightError": 2.0765920538678984e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005536414682865143, + "meanSquaredWeightError": 1.1359857943205043e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010246187448501587, + "meanSquaredWeightError": 7.989317651890815e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014816643670201302, + "meanSquaredWeightError": 6.387295137244564e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009112022817134857, + "meanSquaredWeightError": 3.856152730463691e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0225830078125, + "meanSquaredWeightError": 8.832107027956226e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02226562611758709, + "meanSquaredWeightError": 1.4517542012981721e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.017578125, + "meanSquaredWeightError": 9.189842558043892e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 254279680, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009068762883543968, + "meanSquaredWeightError": 1.1976288405435298e-08, + "storageBytes": 286064640 + } + ], + "quantizationTiers": [ + { + "scheme": "symmetric-int8", + "method": "rtn", + "groupSize": 32, + "include": [], + "exclude": [], + "matrices": 0 + }, + { + "scheme": "symmetric-int4", + "method": "gptq", + "groupSize": 32, + "include": [ + "p_model_model_language_model_layers_0_mlp_down_proj_weight", + "p_model_model_language_model_layers_0_mlp_gate_proj_weight", + "p_model_model_language_model_layers_0_mlp_up_proj_weight", + "p_model_model_language_model_layers_10_mlp_down_proj_weight", + "p_model_model_language_model_layers_10_mlp_gate_proj_weight", + "p_model_model_language_model_layers_10_mlp_up_proj_weight", + "p_model_model_language_model_layers_11_mlp_down_proj_weight", + "p_model_model_language_model_layers_11_mlp_gate_proj_weight", + "p_model_model_language_model_layers_11_mlp_up_proj_weight", + "p_model_model_language_model_layers_12_mlp_down_proj_weight", + "p_model_model_language_model_layers_12_mlp_gate_proj_weight", + "p_model_model_language_model_layers_12_mlp_up_proj_weight", + "p_model_model_language_model_layers_13_mlp_down_proj_weight", + "p_model_model_language_model_layers_13_mlp_gate_proj_weight", + "p_model_model_language_model_layers_13_mlp_up_proj_weight", + "p_model_model_language_model_layers_14_mlp_down_proj_weight", + "p_model_model_language_model_layers_14_mlp_gate_proj_weight", + "p_model_model_language_model_layers_14_mlp_up_proj_weight", + "p_model_model_language_model_layers_15_mlp_down_proj_weight", + "p_model_model_language_model_layers_15_mlp_gate_proj_weight", + "p_model_model_language_model_layers_15_mlp_up_proj_weight", + "p_model_model_language_model_layers_16_mlp_down_proj_weight", + "p_model_model_language_model_layers_16_mlp_gate_proj_weight", + "p_model_model_language_model_layers_16_mlp_up_proj_weight", + "p_model_model_language_model_layers_17_mlp_down_proj_weight", + "p_model_model_language_model_layers_17_mlp_gate_proj_weight", + "p_model_model_language_model_layers_17_mlp_up_proj_weight", + "p_model_model_language_model_layers_18_mlp_down_proj_weight", + "p_model_model_language_model_layers_18_mlp_gate_proj_weight", + "p_model_model_language_model_layers_18_mlp_up_proj_weight", + "p_model_model_language_model_layers_19_mlp_down_proj_weight", + "p_model_model_language_model_layers_19_mlp_gate_proj_weight", + "p_model_model_language_model_layers_19_mlp_up_proj_weight", + "p_model_model_language_model_layers_1_mlp_down_proj_weight", + "p_model_model_language_model_layers_1_mlp_gate_proj_weight", + "p_model_model_language_model_layers_1_mlp_up_proj_weight", + "p_model_model_language_model_layers_20_mlp_down_proj_weight", + "p_model_model_language_model_layers_20_mlp_gate_proj_weight", + "p_model_model_language_model_layers_20_mlp_up_proj_weight", + "p_model_model_language_model_layers_21_mlp_down_proj_weight", + "p_model_model_language_model_layers_21_mlp_gate_proj_weight", + "p_model_model_language_model_layers_21_mlp_up_proj_weight", + "p_model_model_language_model_layers_22_mlp_down_proj_weight", + "p_model_model_language_model_layers_22_mlp_gate_proj_weight", + "p_model_model_language_model_layers_22_mlp_up_proj_weight", + "p_model_model_language_model_layers_23_mlp_down_proj_weight", + "p_model_model_language_model_layers_23_mlp_gate_proj_weight", + "p_model_model_language_model_layers_23_mlp_up_proj_weight", + "p_model_model_language_model_layers_2_mlp_down_proj_weight", + "p_model_model_language_model_layers_2_mlp_gate_proj_weight", + "p_model_model_language_model_layers_2_mlp_up_proj_weight", + "p_model_model_language_model_layers_3_mlp_down_proj_weight", + "p_model_model_language_model_layers_3_mlp_gate_proj_weight", + "p_model_model_language_model_layers_3_mlp_up_proj_weight", + "p_model_model_language_model_layers_4_mlp_down_proj_weight", + "p_model_model_language_model_layers_4_mlp_gate_proj_weight", + "p_model_model_language_model_layers_4_mlp_up_proj_weight", + "p_model_model_language_model_layers_5_mlp_down_proj_weight", + "p_model_model_language_model_layers_5_mlp_gate_proj_weight", + "p_model_model_language_model_layers_5_mlp_up_proj_weight", + "p_model_model_language_model_layers_6_mlp_down_proj_weight", + "p_model_model_language_model_layers_6_mlp_gate_proj_weight", + "p_model_model_language_model_layers_6_mlp_up_proj_weight", + "p_model_model_language_model_layers_7_mlp_down_proj_weight", + "p_model_model_language_model_layers_7_mlp_gate_proj_weight", + "p_model_model_language_model_layers_7_mlp_up_proj_weight", + "p_model_model_language_model_layers_8_mlp_down_proj_weight", + "p_model_model_language_model_layers_8_mlp_gate_proj_weight", + "p_model_model_language_model_layers_8_mlp_up_proj_weight", + "p_model_model_language_model_layers_9_mlp_down_proj_weight", + "p_model_model_language_model_layers_9_mlp_gate_proj_weight", + "p_model_model_language_model_layers_9_mlp_up_proj_weight" + ], + "exclude": [], + "matrices": 72 + }, + { + "scheme": "symmetric-int8", + "method": "rtn", + "groupSize": 32, + "include": [ + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_11_self_attn_k_proj_weight", + "p_model_model_language_model_layers_11_self_attn_o_proj_weight", + "p_model_model_language_model_layers_11_self_attn_q_proj_weight", + "p_model_model_language_model_layers_11_self_attn_v_proj_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_15_self_attn_k_proj_weight", + "p_model_model_language_model_layers_15_self_attn_o_proj_weight", + "p_model_model_language_model_layers_15_self_attn_q_proj_weight", + "p_model_model_language_model_layers_15_self_attn_v_proj_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_19_self_attn_k_proj_weight", + "p_model_model_language_model_layers_19_self_attn_o_proj_weight", + "p_model_model_language_model_layers_19_self_attn_q_proj_weight", + "p_model_model_language_model_layers_19_self_attn_v_proj_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_23_self_attn_k_proj_weight", + "p_model_model_language_model_layers_23_self_attn_o_proj_weight", + "p_model_model_language_model_layers_23_self_attn_q_proj_weight", + "p_model_model_language_model_layers_23_self_attn_v_proj_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_3_self_attn_k_proj_weight", + "p_model_model_language_model_layers_3_self_attn_o_proj_weight", + "p_model_model_language_model_layers_3_self_attn_q_proj_weight", + "p_model_model_language_model_layers_3_self_attn_v_proj_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_7_self_attn_k_proj_weight", + "p_model_model_language_model_layers_7_self_attn_o_proj_weight", + "p_model_model_language_model_layers_7_self_attn_q_proj_weight", + "p_model_model_language_model_layers_7_self_attn_v_proj_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight" + ], + "exclude": [], + "matrices": 114 + } + ], + "states": [ + { + "input": "states_0", + "output": "slice_10", + "kind": "convolution" + }, + { + "input": "states_1", + "output": "slice_16", + "kind": "recurrent" + }, + { + "input": "states_2", + "output": "slice_18", + "kind": "convolution" + }, + { + "input": "states_3", + "output": "slice_24", + "kind": "recurrent" + }, + { + "input": "states_4", + "output": "slice_26", + "kind": "convolution" + }, + { + "input": "states_5", + "output": "slice_32", + "kind": "recurrent" + }, + { + "input": "states_6", + "output": "cache_update", + "kind": "kv" + }, + { + "input": "states_7", + "output": "cache_update_1", + "kind": "kv" + }, + { + "input": "states_8", + "output": "slice_44", + "kind": "convolution" + }, + { + "input": "states_9", + "output": "slice_50", + "kind": "recurrent" + }, + { + "input": "states_10", + "output": "slice_52", + "kind": "convolution" + }, + { + "input": "states_11", + "output": "slice_58", + "kind": "recurrent" + }, + { + "input": "states_12", + "output": "slice_60", + "kind": "convolution" + }, + { + "input": "states_13", + "output": "slice_66", + "kind": "recurrent" + }, + { + "input": "states_14", + "output": "cache_update_2", + "kind": "kv" + }, + { + "input": "states_15", + "output": "cache_update_3", + "kind": "kv" + }, + { + "input": "states_16", + "output": "slice_78", + "kind": "convolution" + }, + { + "input": "states_17", + "output": "slice_84", + "kind": "recurrent" + }, + { + "input": "states_18", + "output": "slice_86", + "kind": "convolution" + }, + { + "input": "states_19", + "output": "slice_92", + "kind": "recurrent" + }, + { + "input": "states_20", + "output": "slice_94", + "kind": "convolution" + }, + { + "input": "states_21", + "output": "slice_100", + "kind": "recurrent" + }, + { + "input": "states_22", + "output": "cache_update_4", + "kind": "kv" + }, + { + "input": "states_23", + "output": "cache_update_5", + "kind": "kv" + }, + { + "input": "states_24", + "output": "slice_112", + "kind": "convolution" + }, + { + "input": "states_25", + "output": "slice_118", + "kind": "recurrent" + }, + { + "input": "states_26", + "output": "slice_120", + "kind": "convolution" + }, + { + "input": "states_27", + "output": "slice_126", + "kind": "recurrent" + }, + { + "input": "states_28", + "output": "slice_128", + "kind": "convolution" + }, + { + "input": "states_29", + "output": "slice_134", + "kind": "recurrent" + }, + { + "input": "states_30", + "output": "cache_update_6", + "kind": "kv" + }, + { + "input": "states_31", + "output": "cache_update_7", + "kind": "kv" + }, + { + "input": "states_32", + "output": "slice_146", + "kind": "convolution" + }, + { + "input": "states_33", + "output": "slice_152", + "kind": "recurrent" + }, + { + "input": "states_34", + "output": "slice_154", + "kind": "convolution" + }, + { + "input": "states_35", + "output": "slice_160", + "kind": "recurrent" + }, + { + "input": "states_36", + "output": "slice_162", + "kind": "convolution" + }, + { + "input": "states_37", + "output": "slice_168", + "kind": "recurrent" + }, + { + "input": "states_38", + "output": "cache_update_8", + "kind": "kv" + }, + { + "input": "states_39", + "output": "cache_update_9", + "kind": "kv" + }, + { + "input": "states_40", + "output": "slice_180", + "kind": "convolution" + }, + { + "input": "states_41", + "output": "slice_186", + "kind": "recurrent" + }, + { + "input": "states_42", + "output": "slice_188", + "kind": "convolution" + }, + { + "input": "states_43", + "output": "slice_194", + "kind": "recurrent" + }, + { + "input": "states_44", + "output": "slice_196", + "kind": "convolution" + }, + { + "input": "states_45", + "output": "slice_202", + "kind": "recurrent" + }, + { + "input": "states_46", + "output": "cache_update_10", + "kind": "kv" + }, + { + "input": "states_47", + "output": "cache_update_11", + "kind": "kv" + } + ], + "dispatches": [ + { + "node": "embedding__part_0", + "inputs": [ + "input_ids", + "p_model_model_language_model_embed_tokens_weight__rows_0_131072", + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072" + ], + "output": "embedding__part_0", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/3eef2865950af58196f0e6e4eaef47123acfcbd576a636f5278aad5a998e4ba6.wgsl" + }, + { + "node": "embedding__part_1", + "inputs": [ + "input_ids", + "p_model_model_language_model_embed_tokens_weight__rows_131072_117248", + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248" + ], + "output": "embedding__part_1", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/96924710cde2ae782e03138b4399e2ed2406f8b37c31401f5a79af528c2193fd.wgsl" + }, + { + "node": "embedding", + "inputs": [ + "embedding__part_0", + "embedding__part_1" + ], + "output": "embedding", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "unsqueeze", + "inputs": [ + "position_ids" + ], + "output": "unsqueeze", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/3203e08f2c99fa4848f553739b7115a7c677840344f17b1d4391d6130517f5a7.wgsl" + }, + { + "node": "expand", + "inputs": [ + "unsqueeze" + ], + "output": "expand", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/52b439507af68e969a6e116d3c8ea5338a4cde5a3fc1e97b729512d17ab2f4f9.wgsl" + }, + { + "node": "slice_1", + "inputs": [ + "expand" + ], + "output": "slice_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/11a92e9cba6bfc381c98c5127db137639d6c87f1cb220459880ead039dc81256.wgsl" + }, + { + "node": "expand_1", + "inputs": [ + "b_model_model_language_model_rotary_emb_inv_freq" + ], + "output": "expand_1", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/a77f14c89662d2d3da8abd8f109667793dc662d307e5e1c61123be753e25437b.wgsl" + }, + { + "node": "_to_copy", + "inputs": [ + "slice_1" + ], + "output": "_to_copy", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/a2bd65fc2709e093644d6cbf0c292b4d71520009a87bedcf4bca8fcb5f5c15b8.wgsl" + }, + { + "node": "matmul", + "inputs": [ + "expand_1", + "_to_copy" + ], + "output": "matmul", + "workgroups": [ + 6, + 1, + 1 + ], + "kernel": "kernels/a87152b0fd2f78f60d7abb61c2f6c4a9bde8a9dd45d0275b58d760a3da9f71ce.wgsl" + }, + { + "node": "transpose", + "inputs": [ + "matmul" + ], + "output": "transpose", + "workgroups": [ + 6, + 1, + 1 + ], + "kernel": "kernels/0f943f386d2f37a8e56b4285b26e1c093c2f2fa2a5ce81cc43c44cc9bb0137f5.wgsl" + }, + { + "node": "select", + "inputs": [ + "transpose" + ], + "output": "select", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/95cba13b595301afbefe20da4a65b93d4d5c5f4a29b9e1d5348d32cb193c6b4b.wgsl" + }, + { + "node": "select_1", + "inputs": [ + "transpose" + ], + "output": "select_1", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/8cce7bfa491959f6a7b62ae9cb861d12e2f4ff05227918a4d06bd80669f0a539.wgsl" + }, + { + "node": "slice_2", + "inputs": [ + "select_1" + ], + "output": "slice_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/18b1cef2bae996f56664d9215180e3bfc77b4e514db52fdf9b6c58be51b53edb.wgsl" + }, + { + "node": "slice_3", + "inputs": [ + "select" + ], + "output": "slice_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/18b1cef2bae996f56664d9215180e3bfc77b4e514db52fdf9b6c58be51b53edb.wgsl" + }, + { + "node": "copy", + "inputs": [ + "slice_2" + ], + "output": "copy", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/0725908abf78c022f86deb338a06779f6f1346ebfe54ff5a43ee6b9eeb93c972.wgsl" + }, + { + "node": "transpose_1", + "inputs": [ + "matmul" + ], + "output": "transpose_1", + "workgroups": [ + 6, + 1, + 1 + ], + "kernel": "kernels/0f943f386d2f37a8e56b4285b26e1c093c2f2fa2a5ce81cc43c44cc9bb0137f5.wgsl" + }, + { + "node": "select_2", + "inputs": [ + "transpose_1" + ], + "output": "select_2", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/95cba13b595301afbefe20da4a65b93d4d5c5f4a29b9e1d5348d32cb193c6b4b.wgsl" + }, + { + "node": "slice_scatter", + "inputs": [ + "copy", + "select_2" + ], + "output": "slice_scatter", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/1545025e0a625ccc73b42ae112c6f9d78a1504c10b759979ae2247984f25f056.wgsl" + }, + { + "node": "select_scatter", + "inputs": [ + "slice_scatter", + "transpose_1" + ], + "output": "select_scatter", + "workgroups": [ + 6, + 1, + 1 + ], + "kernel": "kernels/3f89d19bd6791657fa702ad00a0517b5f292dbf7d0c196bd5b21c949476d23fd.wgsl" + }, + { + "node": "transpose_2", + "inputs": [ + "select_scatter" + ], + "output": "transpose_2", + "workgroups": [ + 6, + 1, + 1 + ], + "kernel": "kernels/2cd4a533d32097a17641196f0cd315d170c587bc2b1c62b42b79d566d6fbf140.wgsl" + }, + { + "node": "transpose_7", + "inputs": [ + "transpose_2" + ], + "output": "transpose_7", + "workgroups": [ + 6, + 1, + 1 + ], + "kernel": "kernels/0f943f386d2f37a8e56b4285b26e1c093c2f2fa2a5ce81cc43c44cc9bb0137f5.wgsl" + }, + { + "node": "select_7", + "inputs": [ + "transpose_7" + ], + "output": "select_7", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/95cba13b595301afbefe20da4a65b93d4d5c5f4a29b9e1d5348d32cb193c6b4b.wgsl" + }, + { + "node": "slice_7", + "inputs": [ + "select_7" + ], + "output": "slice_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/752da00e33abf87b63eba03893a16e7db848f97f1385942d40d8cfb5122d9bd3.wgsl" + }, + { + "node": "transpose_8", + "inputs": [ + "transpose_2" + ], + "output": "transpose_8", + "workgroups": [ + 6, + 1, + 1 + ], + "kernel": "kernels/0f943f386d2f37a8e56b4285b26e1c093c2f2fa2a5ce81cc43c44cc9bb0137f5.wgsl" + }, + { + "node": "select_8", + "inputs": [ + "transpose_8" + ], + "output": "select_8", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/62f30a62c18ed78c9bf37bdc56529accbee2264b9eac931ba0b35a80bb917e7c.wgsl" + }, + { + "node": "slice_8", + "inputs": [ + "select_8" + ], + "output": "slice_8", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/752da00e33abf87b63eba03893a16e7db848f97f1385942d40d8cfb5122d9bd3.wgsl" + }, + { + "node": "copy_1", + "inputs": [ + "slice_8" + ], + "output": "copy_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/7928d1b13d79c16f6c0f9d3130cbdb4193cab6efe3a60b38caa1039688fe56de.wgsl" + }, + { + "node": "transpose_9", + "inputs": [ + "transpose_2" + ], + "output": "transpose_9", + "workgroups": [ + 6, + 1, + 1 + ], + "kernel": "kernels/0f943f386d2f37a8e56b4285b26e1c093c2f2fa2a5ce81cc43c44cc9bb0137f5.wgsl" + }, + { + "node": "select_9", + "inputs": [ + "transpose_9" + ], + "output": "select_9", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/95cba13b595301afbefe20da4a65b93d4d5c5f4a29b9e1d5348d32cb193c6b4b.wgsl" + }, + { + "node": "slice_scatter_1", + "inputs": [ + "copy_1", + "select_9" + ], + "output": "slice_scatter_1", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/ae69e698c92571766e36ad8c078b13e42bde56745d45698daaadaca587b7279c.wgsl" + }, + { + "node": "select_scatter_1", + "inputs": [ + "slice_scatter_1", + "transpose_9" + ], + "output": "select_scatter_1", + "workgroups": [ + 6, + 1, + 1 + ], + "kernel": "kernels/3f89d19bd6791657fa702ad00a0517b5f292dbf7d0c196bd5b21c949476d23fd.wgsl" + }, + { + "node": "transpose_10", + "inputs": [ + "select_scatter_1" + ], + "output": "transpose_10", + "workgroups": [ + 6, + 1, + 1 + ], + "kernel": "kernels/2cd4a533d32097a17641196f0cd315d170c587bc2b1c62b42b79d566d6fbf140.wgsl" + }, + { + "node": "transpose_12", + "inputs": [ + "transpose_10" + ], + "output": "transpose_12", + "workgroups": [ + 6, + 1, + 1 + ], + "kernel": "kernels/0f943f386d2f37a8e56b4285b26e1c093c2f2fa2a5ce81cc43c44cc9bb0137f5.wgsl" + }, + { + "node": "select_11", + "inputs": [ + "transpose_12" + ], + "output": "select_11", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/95cba13b595301afbefe20da4a65b93d4d5c5f4a29b9e1d5348d32cb193c6b4b.wgsl" + }, + { + "node": "cat", + "inputs": [ + "select_11" + ], + "output": "cat", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/0d3892883d1532d173a7dbe85d9ab1daa87f0d9b4923a535b8d2003f24f9500c.wgsl" + }, + { + "node": "cos", + "inputs": [ + "cat" + ], + "output": "cos", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/d8c79a22e83e4a3b75b732bffa9ca673713bb57d580c5af4e59c805645ff1ca3.wgsl" + }, + { + "node": "mul", + "inputs": [ + "cos" + ], + "output": "mul", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/e84c817e08236e486418f6b0b81f5abd27756f82be67cfbc15736410ac3c7e5d.wgsl" + }, + { + "node": "sin", + "inputs": [ + "cat" + ], + "output": "sin", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/0b8815b1efe176d55ab08530c0a9e1913cf6c9ef3e773c574843573d237cf196.wgsl" + }, + { + "node": "mul_1", + "inputs": [ + "sin" + ], + "output": "mul_1", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/e84c817e08236e486418f6b0b81f5abd27756f82be67cfbc15736410ac3c7e5d.wgsl" + }, + { + "node": "mul_3", + "inputs": [ + "embedding", + "p_model_model_language_model_layers_0_input_layernorm_weight" + ], + "output": "mul_3", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear", + "inputs": [ + "mul_3", + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_13", + "inputs": [ + "linear" + ], + "output": "transpose_13", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_1", + "inputs": [ + "mul_3", + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_1", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_2", + "inputs": [ + "mul_3", + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_3", + "inputs": [ + "mul_3", + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_1", + "inputs": [ + "states_0", + "transpose_13" + ], + "output": "cat_1", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_10", + "inputs": [ + "cat_1" + ], + "output": "slice_10", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_0" + }, + { + "node": "conv1d", + "inputs": [ + "cat_1", + "p_model_model_language_model_layers_0_linear_attn_conv1d_weight" + ], + "output": "conv1d", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_11", + "inputs": [ + "conv1d" + ], + "output": "slice_11", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu", + "inputs": [ + "slice_11" + ], + "output": "silu", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_12", + "inputs": [ + "silu" + ], + "output": "slice_12", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_14", + "inputs": [ + "slice_12" + ], + "output": "transpose_14", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_13", + "inputs": [ + "transpose_14" + ], + "output": "slice_13", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_14", + "inputs": [ + "transpose_14" + ], + "output": "slice_14", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_15", + "inputs": [ + "transpose_14" + ], + "output": "slice_15", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid", + "inputs": [ + "linear_2" + ], + "output": "sigmoid", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp", + "inputs": [ + "p_model_model_language_model_layers_0_linear_attn_a_log" + ], + "output": "exp", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg", + "inputs": [ + "exp" + ], + "output": "neg", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_2", + "inputs": [ + "linear_3", + "p_model_model_language_model_layers_0_linear_attn_dt_bias" + ], + "output": "add_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus", + "inputs": [ + "add_2" + ], + "output": "softplus", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_4", + "inputs": [ + "neg", + "softplus" + ], + "output": "mul_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_5", + "inputs": [ + "slice_13" + ], + "output": "mul_5", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_1", + "inputs": [ + "mul_5" + ], + "output": "sum_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_3", + "inputs": [ + "sum_1" + ], + "output": "add_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_1", + "inputs": [ + "add_3" + ], + "output": "rsqrt_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_6", + "inputs": [ + "slice_13", + "rsqrt_1" + ], + "output": "mul_6", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_7", + "inputs": [ + "slice_14" + ], + "output": "mul_7", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_2", + "inputs": [ + "mul_7" + ], + "output": "sum_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_4", + "inputs": [ + "sum_2" + ], + "output": "add_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_2", + "inputs": [ + "add_4" + ], + "output": "rsqrt_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_8", + "inputs": [ + "slice_14", + "rsqrt_2" + ], + "output": "mul_8", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule", + "inputs": [ + "mul_6", + "mul_8", + "slice_15", + "mul_4", + "sigmoid", + "states_1" + ], + "output": "normalized_gated_delta_rule", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_16", + "inputs": [ + "normalized_gated_delta_rule" + ], + "output": "slice_16", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_1" + }, + { + "node": "slice_17", + "inputs": [ + "normalized_gated_delta_rule" + ], + "output": "slice_17", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_15", + "inputs": [ + "slice_17" + ], + "output": "transpose_15", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_10", + "inputs": [ + "transpose_15", + "p_model_model_language_model_layers_0_linear_attn_norm_weight" + ], + "output": "mul_10", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_11", + "inputs": [ + "linear_1", + "mul_10" + ], + "output": "mul_11", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_4", + "inputs": [ + "mul_11", + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_4", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_6", + "inputs": [ + "embedding", + "linear_4" + ], + "output": "add_6", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_13", + "inputs": [ + "add_6", + "p_model_model_language_model_layers_0_post_attention_layernorm_weight" + ], + "output": "mul_13", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_5", + "inputs": [ + "mul_13", + "p_model_model_language_model_layers_0_mlp_gate_proj_weight", + "p_model_model_language_model_layers_0_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_5", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_6", + "inputs": [ + "mul_13", + "p_model_model_language_model_layers_0_mlp_up_proj_weight", + "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_6", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_14", + "inputs": [ + "linear_5", + "linear_6" + ], + "output": "mul_14", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_7", + "inputs": [ + "mul_14", + "p_model_model_language_model_layers_0_mlp_down_proj_weight", + "p_model_model_language_model_layers_0_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_7", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_9", + "inputs": [ + "add_6", + "linear_7" + ], + "output": "add_9", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_16", + "inputs": [ + "add_9", + "p_model_model_language_model_layers_1_input_layernorm_weight" + ], + "output": "mul_16", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_8", + "inputs": [ + "mul_16", + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_8", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_16", + "inputs": [ + "linear_8" + ], + "output": "transpose_16", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_9", + "inputs": [ + "mul_16", + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_9", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_10", + "inputs": [ + "mul_16", + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_11", + "inputs": [ + "mul_16", + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_2", + "inputs": [ + "states_2", + "transpose_16" + ], + "output": "cat_2", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_18", + "inputs": [ + "cat_2" + ], + "output": "slice_18", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_2" + }, + { + "node": "conv1d_1", + "inputs": [ + "cat_2", + "p_model_model_language_model_layers_1_linear_attn_conv1d_weight" + ], + "output": "conv1d_1", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_19", + "inputs": [ + "conv1d_1" + ], + "output": "slice_19", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_3", + "inputs": [ + "slice_19" + ], + "output": "silu_3", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_20", + "inputs": [ + "silu_3" + ], + "output": "slice_20", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_17", + "inputs": [ + "slice_20" + ], + "output": "transpose_17", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_21", + "inputs": [ + "transpose_17" + ], + "output": "slice_21", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_22", + "inputs": [ + "transpose_17" + ], + "output": "slice_22", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_23", + "inputs": [ + "transpose_17" + ], + "output": "slice_23", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_1", + "inputs": [ + "linear_10" + ], + "output": "sigmoid_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_1", + "inputs": [ + "p_model_model_language_model_layers_1_linear_attn_a_log" + ], + "output": "exp_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_1", + "inputs": [ + "exp_1" + ], + "output": "neg_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_12", + "inputs": [ + "linear_11", + "p_model_model_language_model_layers_1_linear_attn_dt_bias" + ], + "output": "add_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_1", + "inputs": [ + "add_12" + ], + "output": "softplus_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_17", + "inputs": [ + "neg_1", + "softplus_1" + ], + "output": "mul_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_18", + "inputs": [ + "slice_21" + ], + "output": "mul_18", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_3", + "inputs": [ + "mul_18" + ], + "output": "sum_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_13", + "inputs": [ + "sum_3" + ], + "output": "add_13", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_6", + "inputs": [ + "add_13" + ], + "output": "rsqrt_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_19", + "inputs": [ + "slice_21", + "rsqrt_6" + ], + "output": "mul_19", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_20", + "inputs": [ + "slice_22" + ], + "output": "mul_20", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_4", + "inputs": [ + "mul_20" + ], + "output": "sum_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_14", + "inputs": [ + "sum_4" + ], + "output": "add_14", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_7", + "inputs": [ + "add_14" + ], + "output": "rsqrt_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_21", + "inputs": [ + "slice_22", + "rsqrt_7" + ], + "output": "mul_21", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_1", + "inputs": [ + "mul_19", + "mul_21", + "slice_23", + "mul_17", + "sigmoid_1", + "states_3" + ], + "output": "normalized_gated_delta_rule_1", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_24", + "inputs": [ + "normalized_gated_delta_rule_1" + ], + "output": "slice_24", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_3" + }, + { + "node": "slice_25", + "inputs": [ + "normalized_gated_delta_rule_1" + ], + "output": "slice_25", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_18", + "inputs": [ + "slice_25" + ], + "output": "transpose_18", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_23", + "inputs": [ + "transpose_18", + "p_model_model_language_model_layers_1_linear_attn_norm_weight" + ], + "output": "mul_23", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_24", + "inputs": [ + "linear_9", + "mul_23" + ], + "output": "mul_24", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_12", + "inputs": [ + "mul_24", + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_12", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_16", + "inputs": [ + "add_9", + "linear_12" + ], + "output": "add_16", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_26", + "inputs": [ + "add_16", + "p_model_model_language_model_layers_1_post_attention_layernorm_weight" + ], + "output": "mul_26", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_13", + "inputs": [ + "mul_26", + "p_model_model_language_model_layers_1_mlp_gate_proj_weight", + "p_model_model_language_model_layers_1_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_13", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_14", + "inputs": [ + "mul_26", + "p_model_model_language_model_layers_1_mlp_up_proj_weight", + "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_14", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_27", + "inputs": [ + "linear_13", + "linear_14" + ], + "output": "mul_27", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_15", + "inputs": [ + "mul_27", + "p_model_model_language_model_layers_1_mlp_down_proj_weight", + "p_model_model_language_model_layers_1_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_15", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_19", + "inputs": [ + "add_16", + "linear_15" + ], + "output": "add_19", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_29", + "inputs": [ + "add_19", + "p_model_model_language_model_layers_2_input_layernorm_weight" + ], + "output": "mul_29", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_16", + "inputs": [ + "mul_29", + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_16", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_19", + "inputs": [ + "linear_16" + ], + "output": "transpose_19", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_17", + "inputs": [ + "mul_29", + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_17", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_18", + "inputs": [ + "mul_29", + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_18", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_19", + "inputs": [ + "mul_29", + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_19", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_3", + "inputs": [ + "states_4", + "transpose_19" + ], + "output": "cat_3", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_26", + "inputs": [ + "cat_3" + ], + "output": "slice_26", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_4" + }, + { + "node": "conv1d_2", + "inputs": [ + "cat_3", + "p_model_model_language_model_layers_2_linear_attn_conv1d_weight" + ], + "output": "conv1d_2", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_27", + "inputs": [ + "conv1d_2" + ], + "output": "slice_27", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_6", + "inputs": [ + "slice_27" + ], + "output": "silu_6", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_28", + "inputs": [ + "silu_6" + ], + "output": "slice_28", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_20", + "inputs": [ + "slice_28" + ], + "output": "transpose_20", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_29", + "inputs": [ + "transpose_20" + ], + "output": "slice_29", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_30", + "inputs": [ + "transpose_20" + ], + "output": "slice_30", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_31", + "inputs": [ + "transpose_20" + ], + "output": "slice_31", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_2", + "inputs": [ + "linear_18" + ], + "output": "sigmoid_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_2", + "inputs": [ + "p_model_model_language_model_layers_2_linear_attn_a_log" + ], + "output": "exp_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_2", + "inputs": [ + "exp_2" + ], + "output": "neg_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_22", + "inputs": [ + "linear_19", + "p_model_model_language_model_layers_2_linear_attn_dt_bias" + ], + "output": "add_22", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_2", + "inputs": [ + "add_22" + ], + "output": "softplus_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_30", + "inputs": [ + "neg_2", + "softplus_2" + ], + "output": "mul_30", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_31", + "inputs": [ + "slice_29" + ], + "output": "mul_31", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_5", + "inputs": [ + "mul_31" + ], + "output": "sum_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_23", + "inputs": [ + "sum_5" + ], + "output": "add_23", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_11", + "inputs": [ + "add_23" + ], + "output": "rsqrt_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_32", + "inputs": [ + "slice_29", + "rsqrt_11" + ], + "output": "mul_32", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_33", + "inputs": [ + "slice_30" + ], + "output": "mul_33", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_6", + "inputs": [ + "mul_33" + ], + "output": "sum_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_24", + "inputs": [ + "sum_6" + ], + "output": "add_24", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_12", + "inputs": [ + "add_24" + ], + "output": "rsqrt_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_34", + "inputs": [ + "slice_30", + "rsqrt_12" + ], + "output": "mul_34", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_2", + "inputs": [ + "mul_32", + "mul_34", + "slice_31", + "mul_30", + "sigmoid_2", + "states_5" + ], + "output": "normalized_gated_delta_rule_2", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_32", + "inputs": [ + "normalized_gated_delta_rule_2" + ], + "output": "slice_32", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_5" + }, + { + "node": "slice_33", + "inputs": [ + "normalized_gated_delta_rule_2" + ], + "output": "slice_33", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_21", + "inputs": [ + "slice_33" + ], + "output": "transpose_21", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_36", + "inputs": [ + "transpose_21", + "p_model_model_language_model_layers_2_linear_attn_norm_weight" + ], + "output": "mul_36", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_37", + "inputs": [ + "linear_17", + "mul_36" + ], + "output": "mul_37", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_20", + "inputs": [ + "mul_37", + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_20", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_26", + "inputs": [ + "add_19", + "linear_20" + ], + "output": "add_26", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_39", + "inputs": [ + "add_26", + "p_model_model_language_model_layers_2_post_attention_layernorm_weight" + ], + "output": "mul_39", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_21", + "inputs": [ + "mul_39", + "p_model_model_language_model_layers_2_mlp_gate_proj_weight", + "p_model_model_language_model_layers_2_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_21", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_22", + "inputs": [ + "mul_39", + "p_model_model_language_model_layers_2_mlp_up_proj_weight", + "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_22", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_40", + "inputs": [ + "linear_21", + "linear_22" + ], + "output": "mul_40", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_23", + "inputs": [ + "mul_40", + "p_model_model_language_model_layers_2_mlp_down_proj_weight", + "p_model_model_language_model_layers_2_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_23", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_29", + "inputs": [ + "add_26", + "linear_23" + ], + "output": "add_29", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_42", + "inputs": [ + "add_29", + "p_model_model_language_model_layers_3_input_layernorm_weight" + ], + "output": "mul_42", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_24", + "inputs": [ + "mul_42", + "p_model_model_language_model_layers_3_self_attn_q_proj_weight", + "p_model_model_language_model_layers_3_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_24", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/f7f0ee18ecdc712d36e487eef201cb442f28f824e5b6fac7e4d16cb6b7debec4.wgsl" + }, + { + "node": "slice_34", + "inputs": [ + "linear_24" + ], + "output": "slice_34", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2eeb54034621ff097879064a60fd4e0dd5d68201e48296ad234413a39579acc8.wgsl" + }, + { + "node": "slice_35", + "inputs": [ + "linear_24" + ], + "output": "slice_35", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0d1edb63bf4dc212d55f8a4c79a50e04c0d822bd9f6e193003e2071ba0d57d69.wgsl" + }, + { + "node": "mul_44", + "inputs": [ + "slice_34", + "p_model_model_language_model_layers_3_self_attn_q_norm_weight" + ], + "output": "mul_44", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl" + }, + { + "node": "transpose_22", + "inputs": [ + "mul_44" + ], + "output": "transpose_22", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/18a0b02a12ebf73203283d267342582d54ba411aeee9cbe21ad89ae890e4cfea.wgsl" + }, + { + "node": "linear_25", + "inputs": [ + "mul_42", + "p_model_model_language_model_layers_3_self_attn_k_proj_weight", + "p_model_model_language_model_layers_3_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_25", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl" + }, + { + "node": "mul_46", + "inputs": [ + "linear_25", + "p_model_model_language_model_layers_3_self_attn_k_norm_weight" + ], + "output": "mul_46", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/f994142054bcb47d6a554524dd8de9c3054467a1d07011027c47f475d0909c19.wgsl" + }, + { + "node": "transpose_23", + "inputs": [ + "mul_46" + ], + "output": "transpose_23", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl" + }, + { + "node": "linear_26", + "inputs": [ + "mul_42", + "p_model_model_language_model_layers_3_self_attn_v_proj_weight", + "p_model_model_language_model_layers_3_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_26", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl" + }, + { + "node": "transpose_24", + "inputs": [ + "linear_26" + ], + "output": "transpose_24", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl" + }, + { + "node": "slice_36", + "inputs": [ + "transpose_22" + ], + "output": "slice_36", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8b5e4c9b7889f2bccebf60987b6f5b8c23542c3475cd8d78656dc7ecffd81f8d.wgsl" + }, + { + "node": "slice_37", + "inputs": [ + "transpose_22" + ], + "output": "slice_37", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/f95724b4a3d2dbafaccd23a45f68234bb49e4a39168942060ca628960993f33d.wgsl" + }, + { + "node": "slice_38", + "inputs": [ + "transpose_23" + ], + "output": "slice_38", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/ad72f60d58133272228319f1c066e7cd43ac28573ddeeb7a6f69be01af341f57.wgsl" + }, + { + "node": "slice_39", + "inputs": [ + "transpose_23" + ], + "output": "slice_39", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/8af32f59dfb78c79ac2187ae707f0897b89fb8bb0e3d832e2edac308eefa603c.wgsl" + }, + { + "node": "mul_47", + "inputs": [ + "slice_36", + "mul" + ], + "output": "mul_47", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl" + }, + { + "node": "slice_40", + "inputs": [ + "slice_36" + ], + "output": "slice_40", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8ee286e06c1271162bea74e14b4bd5b1d66890648217c79435e6b3f1d4181ec8.wgsl" + }, + { + "node": "slice_41", + "inputs": [ + "slice_36" + ], + "output": "slice_41", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/80d22263b082805bc9d4499247f92b7a1a783a3e1cdd9c652d304492a1cfd425.wgsl" + }, + { + "node": "neg_3", + "inputs": [ + "slice_41" + ], + "output": "neg_3", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl" + }, + { + "node": "cat_4", + "inputs": [ + "neg_3", + "slice_40" + ], + "output": "cat_4", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/bda7ebd3bcf25c31938fca1a582e82b972ab742f85f826f085417b0e6f651e60.wgsl" + }, + { + "node": "mul_48", + "inputs": [ + "cat_4", + "mul_1" + ], + "output": "mul_48", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl" + }, + { + "node": "add_36", + "inputs": [ + "mul_47", + "mul_48" + ], + "output": "add_36", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl" + }, + { + "node": "mul_49", + "inputs": [ + "slice_38", + "mul" + ], + "output": "mul_49", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/d0be84d22b9e3b0cf73729e9a6fad1c6cf7d03fa01b45b8765ed70158578f2a9.wgsl" + }, + { + "node": "slice_42", + "inputs": [ + "slice_38" + ], + "output": "slice_42", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8c3a77238461b0ae13318bad504e1e0779f3f47d1782fe0ca78f20112c031146.wgsl" + }, + { + "node": "slice_43", + "inputs": [ + "slice_38" + ], + "output": "slice_43", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/b2ae1b161c8a6315f8863e47511607027dfe583aac1db4c1626371ddfa0c64e2.wgsl" + }, + { + "node": "neg_4", + "inputs": [ + "slice_43" + ], + "output": "neg_4", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/4fad29e3e337840e25132f10b3e3d55225f727ae13525ff0b0bdf31e03f09d74.wgsl" + }, + { + "node": "cat_5", + "inputs": [ + "neg_4", + "slice_42" + ], + "output": "cat_5", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/898086c99e03b2cdb78e3e5658c3fff202f1c87d09d8da08659bd1ba85e01b09.wgsl" + }, + { + "node": "mul_50", + "inputs": [ + "cat_5", + "mul_1" + ], + "output": "mul_50", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/d0be84d22b9e3b0cf73729e9a6fad1c6cf7d03fa01b45b8765ed70158578f2a9.wgsl" + }, + { + "node": "add_37", + "inputs": [ + "mul_49", + "mul_50" + ], + "output": "add_37", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/f5b51b1e0e52bd87a70a0ca169186ebb5734cbdc2b8f375ba0241485b1df4e66.wgsl" + }, + { + "node": "cat_6", + "inputs": [ + "add_36", + "slice_37" + ], + "output": "cat_6", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl" + }, + { + "node": "cat_7", + "inputs": [ + "add_37", + "slice_39" + ], + "output": "cat_7", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f276cf6656b999339df464b61934487f5c8eb7a0957cf6d902254ab1aa2ffcd0.wgsl" + }, + { + "node": "cache_update", + "inputs": [ + "cat_7", + "position_ids" + ], + "output": "cache_update", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_6", + "kernel": "kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl" + }, + { + "node": "cache_update_1", + "inputs": [ + "transpose_24", + "position_ids" + ], + "output": "cache_update_1", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_7", + "kernel": "kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl" + }, + { + "node": "matmul_2", + "inputs": [ + "cat_6", + "cache_update", + "cache_update_1", + "attention_mask" + ], + "output": "matmul_2", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f5b2ba528c9d245492ad94b337e8cafcc65deeb4a5e7975a55ca79cbfe0a29d4.wgsl" + }, + { + "node": "transpose_26", + "inputs": [ + "matmul_2" + ], + "output": "transpose_26", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2792ae5841828bd5af938c10096abe385e63bb3ec2a22e850f146739433c12bf.wgsl" + }, + { + "node": "sigmoid_3", + "inputs": [ + "slice_35" + ], + "output": "sigmoid_3", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a53c1ce51ebaec593ad667efae9603efe3807db8f0ee060a620ac031cc4f2b92.wgsl" + }, + { + "node": "mul_52", + "inputs": [ + "transpose_26", + "sigmoid_3" + ], + "output": "mul_52", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/45a1fd3ed514f208718651026bf2ee30dda66b02cdd950dc71e88eb6bdac0b10.wgsl" + }, + { + "node": "linear_27", + "inputs": [ + "mul_52", + "p_model_model_language_model_layers_3_self_attn_o_proj_weight", + "p_model_model_language_model_layers_3_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_27", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_39", + "inputs": [ + "add_29", + "linear_27" + ], + "output": "add_39", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_54", + "inputs": [ + "add_39", + "p_model_model_language_model_layers_3_post_attention_layernorm_weight" + ], + "output": "mul_54", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_28", + "inputs": [ + "mul_54", + "p_model_model_language_model_layers_3_mlp_gate_proj_weight", + "p_model_model_language_model_layers_3_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_28", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_29", + "inputs": [ + "mul_54", + "p_model_model_language_model_layers_3_mlp_up_proj_weight", + "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_29", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_55", + "inputs": [ + "linear_28", + "linear_29" + ], + "output": "mul_55", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_30", + "inputs": [ + "mul_55", + "p_model_model_language_model_layers_3_mlp_down_proj_weight", + "p_model_model_language_model_layers_3_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_30", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_42", + "inputs": [ + "add_39", + "linear_30" + ], + "output": "add_42", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_57", + "inputs": [ + "add_42", + "p_model_model_language_model_layers_4_input_layernorm_weight" + ], + "output": "mul_57", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_31", + "inputs": [ + "mul_57", + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_31", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_27", + "inputs": [ + "linear_31" + ], + "output": "transpose_27", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_32", + "inputs": [ + "mul_57", + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_32", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_33", + "inputs": [ + "mul_57", + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_33", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_34", + "inputs": [ + "mul_57", + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_34", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_8", + "inputs": [ + "states_8", + "transpose_27" + ], + "output": "cat_8", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_44", + "inputs": [ + "cat_8" + ], + "output": "slice_44", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_8" + }, + { + "node": "conv1d_3", + "inputs": [ + "cat_8", + "p_model_model_language_model_layers_4_linear_attn_conv1d_weight" + ], + "output": "conv1d_3", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_45", + "inputs": [ + "conv1d_3" + ], + "output": "slice_45", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_10", + "inputs": [ + "slice_45" + ], + "output": "silu_10", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_46", + "inputs": [ + "silu_10" + ], + "output": "slice_46", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_28", + "inputs": [ + "slice_46" + ], + "output": "transpose_28", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_47", + "inputs": [ + "transpose_28" + ], + "output": "slice_47", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_48", + "inputs": [ + "transpose_28" + ], + "output": "slice_48", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_49", + "inputs": [ + "transpose_28" + ], + "output": "slice_49", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_4", + "inputs": [ + "linear_33" + ], + "output": "sigmoid_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_3", + "inputs": [ + "p_model_model_language_model_layers_4_linear_attn_a_log" + ], + "output": "exp_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_5", + "inputs": [ + "exp_3" + ], + "output": "neg_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_45", + "inputs": [ + "linear_34", + "p_model_model_language_model_layers_4_linear_attn_dt_bias" + ], + "output": "add_45", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_3", + "inputs": [ + "add_45" + ], + "output": "softplus_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_58", + "inputs": [ + "neg_5", + "softplus_3" + ], + "output": "mul_58", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_59", + "inputs": [ + "slice_47" + ], + "output": "mul_59", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_7", + "inputs": [ + "mul_59" + ], + "output": "sum_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_46", + "inputs": [ + "sum_7" + ], + "output": "add_46", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_20", + "inputs": [ + "add_46" + ], + "output": "rsqrt_20", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_60", + "inputs": [ + "slice_47", + "rsqrt_20" + ], + "output": "mul_60", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_61", + "inputs": [ + "slice_48" + ], + "output": "mul_61", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_8", + "inputs": [ + "mul_61" + ], + "output": "sum_8", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_47", + "inputs": [ + "sum_8" + ], + "output": "add_47", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_21", + "inputs": [ + "add_47" + ], + "output": "rsqrt_21", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_62", + "inputs": [ + "slice_48", + "rsqrt_21" + ], + "output": "mul_62", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_3", + "inputs": [ + "mul_60", + "mul_62", + "slice_49", + "mul_58", + "sigmoid_4", + "states_9" + ], + "output": "normalized_gated_delta_rule_3", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_50", + "inputs": [ + "normalized_gated_delta_rule_3" + ], + "output": "slice_50", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_9" + }, + { + "node": "slice_51", + "inputs": [ + "normalized_gated_delta_rule_3" + ], + "output": "slice_51", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_29", + "inputs": [ + "slice_51" + ], + "output": "transpose_29", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_64", + "inputs": [ + "transpose_29", + "p_model_model_language_model_layers_4_linear_attn_norm_weight" + ], + "output": "mul_64", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_65", + "inputs": [ + "linear_32", + "mul_64" + ], + "output": "mul_65", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_35", + "inputs": [ + "mul_65", + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_35", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_49", + "inputs": [ + "add_42", + "linear_35" + ], + "output": "add_49", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_67", + "inputs": [ + "add_49", + "p_model_model_language_model_layers_4_post_attention_layernorm_weight" + ], + "output": "mul_67", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_36", + "inputs": [ + "mul_67", + "p_model_model_language_model_layers_4_mlp_gate_proj_weight", + "p_model_model_language_model_layers_4_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_36", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_37", + "inputs": [ + "mul_67", + "p_model_model_language_model_layers_4_mlp_up_proj_weight", + "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_37", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_68", + "inputs": [ + "linear_36", + "linear_37" + ], + "output": "mul_68", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_38", + "inputs": [ + "mul_68", + "p_model_model_language_model_layers_4_mlp_down_proj_weight", + "p_model_model_language_model_layers_4_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_38", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_52", + "inputs": [ + "add_49", + "linear_38" + ], + "output": "add_52", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_70", + "inputs": [ + "add_52", + "p_model_model_language_model_layers_5_input_layernorm_weight" + ], + "output": "mul_70", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_39", + "inputs": [ + "mul_70", + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_39", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_30", + "inputs": [ + "linear_39" + ], + "output": "transpose_30", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_40", + "inputs": [ + "mul_70", + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_40", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_41", + "inputs": [ + "mul_70", + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_41", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_42", + "inputs": [ + "mul_70", + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_42", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_9", + "inputs": [ + "states_10", + "transpose_30" + ], + "output": "cat_9", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_52", + "inputs": [ + "cat_9" + ], + "output": "slice_52", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_10" + }, + { + "node": "conv1d_4", + "inputs": [ + "cat_9", + "p_model_model_language_model_layers_5_linear_attn_conv1d_weight" + ], + "output": "conv1d_4", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_53", + "inputs": [ + "conv1d_4" + ], + "output": "slice_53", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_13", + "inputs": [ + "slice_53" + ], + "output": "silu_13", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_54", + "inputs": [ + "silu_13" + ], + "output": "slice_54", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_31", + "inputs": [ + "slice_54" + ], + "output": "transpose_31", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_55", + "inputs": [ + "transpose_31" + ], + "output": "slice_55", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_56", + "inputs": [ + "transpose_31" + ], + "output": "slice_56", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_57", + "inputs": [ + "transpose_31" + ], + "output": "slice_57", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_5", + "inputs": [ + "linear_41" + ], + "output": "sigmoid_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_4", + "inputs": [ + "p_model_model_language_model_layers_5_linear_attn_a_log" + ], + "output": "exp_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_6", + "inputs": [ + "exp_4" + ], + "output": "neg_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_55", + "inputs": [ + "linear_42", + "p_model_model_language_model_layers_5_linear_attn_dt_bias" + ], + "output": "add_55", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_4", + "inputs": [ + "add_55" + ], + "output": "softplus_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_71", + "inputs": [ + "neg_6", + "softplus_4" + ], + "output": "mul_71", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_72", + "inputs": [ + "slice_55" + ], + "output": "mul_72", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_9", + "inputs": [ + "mul_72" + ], + "output": "sum_9", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_56", + "inputs": [ + "sum_9" + ], + "output": "add_56", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_25", + "inputs": [ + "add_56" + ], + "output": "rsqrt_25", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_73", + "inputs": [ + "slice_55", + "rsqrt_25" + ], + "output": "mul_73", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_74", + "inputs": [ + "slice_56" + ], + "output": "mul_74", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_10", + "inputs": [ + "mul_74" + ], + "output": "sum_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_57", + "inputs": [ + "sum_10" + ], + "output": "add_57", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_26", + "inputs": [ + "add_57" + ], + "output": "rsqrt_26", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_75", + "inputs": [ + "slice_56", + "rsqrt_26" + ], + "output": "mul_75", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_4", + "inputs": [ + "mul_73", + "mul_75", + "slice_57", + "mul_71", + "sigmoid_5", + "states_11" + ], + "output": "normalized_gated_delta_rule_4", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_58", + "inputs": [ + "normalized_gated_delta_rule_4" + ], + "output": "slice_58", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_11" + }, + { + "node": "slice_59", + "inputs": [ + "normalized_gated_delta_rule_4" + ], + "output": "slice_59", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_32", + "inputs": [ + "slice_59" + ], + "output": "transpose_32", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_77", + "inputs": [ + "transpose_32", + "p_model_model_language_model_layers_5_linear_attn_norm_weight" + ], + "output": "mul_77", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_78", + "inputs": [ + "linear_40", + "mul_77" + ], + "output": "mul_78", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_43", + "inputs": [ + "mul_78", + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_43", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_59", + "inputs": [ + "add_52", + "linear_43" + ], + "output": "add_59", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_80", + "inputs": [ + "add_59", + "p_model_model_language_model_layers_5_post_attention_layernorm_weight" + ], + "output": "mul_80", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_44", + "inputs": [ + "mul_80", + "p_model_model_language_model_layers_5_mlp_gate_proj_weight", + "p_model_model_language_model_layers_5_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_44", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_45", + "inputs": [ + "mul_80", + "p_model_model_language_model_layers_5_mlp_up_proj_weight", + "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_45", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_81", + "inputs": [ + "linear_44", + "linear_45" + ], + "output": "mul_81", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_46", + "inputs": [ + "mul_81", + "p_model_model_language_model_layers_5_mlp_down_proj_weight", + "p_model_model_language_model_layers_5_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_46", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_62", + "inputs": [ + "add_59", + "linear_46" + ], + "output": "add_62", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_83", + "inputs": [ + "add_62", + "p_model_model_language_model_layers_6_input_layernorm_weight" + ], + "output": "mul_83", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_47", + "inputs": [ + "mul_83", + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_47", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_33", + "inputs": [ + "linear_47" + ], + "output": "transpose_33", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_48", + "inputs": [ + "mul_83", + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_48", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_49", + "inputs": [ + "mul_83", + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_49", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_50", + "inputs": [ + "mul_83", + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_50", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_10", + "inputs": [ + "states_12", + "transpose_33" + ], + "output": "cat_10", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_60", + "inputs": [ + "cat_10" + ], + "output": "slice_60", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_12" + }, + { + "node": "conv1d_5", + "inputs": [ + "cat_10", + "p_model_model_language_model_layers_6_linear_attn_conv1d_weight" + ], + "output": "conv1d_5", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_61", + "inputs": [ + "conv1d_5" + ], + "output": "slice_61", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_16", + "inputs": [ + "slice_61" + ], + "output": "silu_16", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_62", + "inputs": [ + "silu_16" + ], + "output": "slice_62", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_34", + "inputs": [ + "slice_62" + ], + "output": "transpose_34", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_63", + "inputs": [ + "transpose_34" + ], + "output": "slice_63", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_64", + "inputs": [ + "transpose_34" + ], + "output": "slice_64", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_65", + "inputs": [ + "transpose_34" + ], + "output": "slice_65", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_6", + "inputs": [ + "linear_49" + ], + "output": "sigmoid_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_5", + "inputs": [ + "p_model_model_language_model_layers_6_linear_attn_a_log" + ], + "output": "exp_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_7", + "inputs": [ + "exp_5" + ], + "output": "neg_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_65", + "inputs": [ + "linear_50", + "p_model_model_language_model_layers_6_linear_attn_dt_bias" + ], + "output": "add_65", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_5", + "inputs": [ + "add_65" + ], + "output": "softplus_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_84", + "inputs": [ + "neg_7", + "softplus_5" + ], + "output": "mul_84", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_85", + "inputs": [ + "slice_63" + ], + "output": "mul_85", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_11", + "inputs": [ + "mul_85" + ], + "output": "sum_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_66", + "inputs": [ + "sum_11" + ], + "output": "add_66", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_30", + "inputs": [ + "add_66" + ], + "output": "rsqrt_30", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_86", + "inputs": [ + "slice_63", + "rsqrt_30" + ], + "output": "mul_86", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_87", + "inputs": [ + "slice_64" + ], + "output": "mul_87", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_12", + "inputs": [ + "mul_87" + ], + "output": "sum_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_67", + "inputs": [ + "sum_12" + ], + "output": "add_67", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_31", + "inputs": [ + "add_67" + ], + "output": "rsqrt_31", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_88", + "inputs": [ + "slice_64", + "rsqrt_31" + ], + "output": "mul_88", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_5", + "inputs": [ + "mul_86", + "mul_88", + "slice_65", + "mul_84", + "sigmoid_6", + "states_13" + ], + "output": "normalized_gated_delta_rule_5", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_66", + "inputs": [ + "normalized_gated_delta_rule_5" + ], + "output": "slice_66", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_13" + }, + { + "node": "slice_67", + "inputs": [ + "normalized_gated_delta_rule_5" + ], + "output": "slice_67", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_35", + "inputs": [ + "slice_67" + ], + "output": "transpose_35", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_90", + "inputs": [ + "transpose_35", + "p_model_model_language_model_layers_6_linear_attn_norm_weight" + ], + "output": "mul_90", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_91", + "inputs": [ + "linear_48", + "mul_90" + ], + "output": "mul_91", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_51", + "inputs": [ + "mul_91", + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_51", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_69", + "inputs": [ + "add_62", + "linear_51" + ], + "output": "add_69", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_93", + "inputs": [ + "add_69", + "p_model_model_language_model_layers_6_post_attention_layernorm_weight" + ], + "output": "mul_93", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_52", + "inputs": [ + "mul_93", + "p_model_model_language_model_layers_6_mlp_gate_proj_weight", + "p_model_model_language_model_layers_6_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_52", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_53", + "inputs": [ + "mul_93", + "p_model_model_language_model_layers_6_mlp_up_proj_weight", + "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_53", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_94", + "inputs": [ + "linear_52", + "linear_53" + ], + "output": "mul_94", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_54", + "inputs": [ + "mul_94", + "p_model_model_language_model_layers_6_mlp_down_proj_weight", + "p_model_model_language_model_layers_6_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_54", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_72", + "inputs": [ + "add_69", + "linear_54" + ], + "output": "add_72", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_96", + "inputs": [ + "add_72", + "p_model_model_language_model_layers_7_input_layernorm_weight" + ], + "output": "mul_96", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_55", + "inputs": [ + "mul_96", + "p_model_model_language_model_layers_7_self_attn_q_proj_weight", + "p_model_model_language_model_layers_7_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_55", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/f7f0ee18ecdc712d36e487eef201cb442f28f824e5b6fac7e4d16cb6b7debec4.wgsl" + }, + { + "node": "slice_68", + "inputs": [ + "linear_55" + ], + "output": "slice_68", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2eeb54034621ff097879064a60fd4e0dd5d68201e48296ad234413a39579acc8.wgsl" + }, + { + "node": "slice_69", + "inputs": [ + "linear_55" + ], + "output": "slice_69", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0d1edb63bf4dc212d55f8a4c79a50e04c0d822bd9f6e193003e2071ba0d57d69.wgsl" + }, + { + "node": "mul_98", + "inputs": [ + "slice_68", + "p_model_model_language_model_layers_7_self_attn_q_norm_weight" + ], + "output": "mul_98", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl" + }, + { + "node": "transpose_36", + "inputs": [ + "mul_98" + ], + "output": "transpose_36", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/18a0b02a12ebf73203283d267342582d54ba411aeee9cbe21ad89ae890e4cfea.wgsl" + }, + { + "node": "linear_56", + "inputs": [ + "mul_96", + "p_model_model_language_model_layers_7_self_attn_k_proj_weight", + "p_model_model_language_model_layers_7_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_56", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl" + }, + { + "node": "mul_100", + "inputs": [ + "linear_56", + "p_model_model_language_model_layers_7_self_attn_k_norm_weight" + ], + "output": "mul_100", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/f994142054bcb47d6a554524dd8de9c3054467a1d07011027c47f475d0909c19.wgsl" + }, + { + "node": "transpose_37", + "inputs": [ + "mul_100" + ], + "output": "transpose_37", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl" + }, + { + "node": "linear_57", + "inputs": [ + "mul_96", + "p_model_model_language_model_layers_7_self_attn_v_proj_weight", + "p_model_model_language_model_layers_7_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_57", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl" + }, + { + "node": "transpose_38", + "inputs": [ + "linear_57" + ], + "output": "transpose_38", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl" + }, + { + "node": "slice_70", + "inputs": [ + "transpose_36" + ], + "output": "slice_70", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8b5e4c9b7889f2bccebf60987b6f5b8c23542c3475cd8d78656dc7ecffd81f8d.wgsl" + }, + { + "node": "slice_71", + "inputs": [ + "transpose_36" + ], + "output": "slice_71", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/f95724b4a3d2dbafaccd23a45f68234bb49e4a39168942060ca628960993f33d.wgsl" + }, + { + "node": "slice_72", + "inputs": [ + "transpose_37" + ], + "output": "slice_72", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/ad72f60d58133272228319f1c066e7cd43ac28573ddeeb7a6f69be01af341f57.wgsl" + }, + { + "node": "slice_73", + "inputs": [ + "transpose_37" + ], + "output": "slice_73", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/8af32f59dfb78c79ac2187ae707f0897b89fb8bb0e3d832e2edac308eefa603c.wgsl" + }, + { + "node": "mul_101", + "inputs": [ + "slice_70", + "mul" + ], + "output": "mul_101", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl" + }, + { + "node": "slice_74", + "inputs": [ + "slice_70" + ], + "output": "slice_74", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8ee286e06c1271162bea74e14b4bd5b1d66890648217c79435e6b3f1d4181ec8.wgsl" + }, + { + "node": "slice_75", + "inputs": [ + "slice_70" + ], + "output": "slice_75", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/80d22263b082805bc9d4499247f92b7a1a783a3e1cdd9c652d304492a1cfd425.wgsl" + }, + { + "node": "neg_8", + "inputs": [ + "slice_75" + ], + "output": "neg_8", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl" + }, + { + "node": "cat_11", + "inputs": [ + "neg_8", + "slice_74" + ], + "output": "cat_11", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/bda7ebd3bcf25c31938fca1a582e82b972ab742f85f826f085417b0e6f651e60.wgsl" + }, + { + "node": "mul_102", + "inputs": [ + "cat_11", + "mul_1" + ], + "output": "mul_102", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl" + }, + { + "node": "add_79", + "inputs": [ + "mul_101", + "mul_102" + ], + "output": "add_79", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl" + }, + { + "node": "mul_103", + "inputs": [ + "slice_72", + "mul" + ], + "output": "mul_103", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/d0be84d22b9e3b0cf73729e9a6fad1c6cf7d03fa01b45b8765ed70158578f2a9.wgsl" + }, + { + "node": "slice_76", + "inputs": [ + "slice_72" + ], + "output": "slice_76", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8c3a77238461b0ae13318bad504e1e0779f3f47d1782fe0ca78f20112c031146.wgsl" + }, + { + "node": "slice_77", + "inputs": [ + "slice_72" + ], + "output": "slice_77", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/b2ae1b161c8a6315f8863e47511607027dfe583aac1db4c1626371ddfa0c64e2.wgsl" + }, + { + "node": "neg_9", + "inputs": [ + "slice_77" + ], + "output": "neg_9", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/4fad29e3e337840e25132f10b3e3d55225f727ae13525ff0b0bdf31e03f09d74.wgsl" + }, + { + "node": "cat_12", + "inputs": [ + "neg_9", + "slice_76" + ], + "output": "cat_12", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/898086c99e03b2cdb78e3e5658c3fff202f1c87d09d8da08659bd1ba85e01b09.wgsl" + }, + { + "node": "mul_104", + "inputs": [ + "cat_12", + "mul_1" + ], + "output": "mul_104", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/d0be84d22b9e3b0cf73729e9a6fad1c6cf7d03fa01b45b8765ed70158578f2a9.wgsl" + }, + { + "node": "add_80", + "inputs": [ + "mul_103", + "mul_104" + ], + "output": "add_80", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/f5b51b1e0e52bd87a70a0ca169186ebb5734cbdc2b8f375ba0241485b1df4e66.wgsl" + }, + { + "node": "cat_13", + "inputs": [ + "add_79", + "slice_71" + ], + "output": "cat_13", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl" + }, + { + "node": "cat_14", + "inputs": [ + "add_80", + "slice_73" + ], + "output": "cat_14", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f276cf6656b999339df464b61934487f5c8eb7a0957cf6d902254ab1aa2ffcd0.wgsl" + }, + { + "node": "cache_update_2", + "inputs": [ + "cat_14", + "position_ids" + ], + "output": "cache_update_2", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_14", + "kernel": "kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl" + }, + { + "node": "cache_update_3", + "inputs": [ + "transpose_38", + "position_ids" + ], + "output": "cache_update_3", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_15", + "kernel": "kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl" + }, + { + "node": "matmul_4", + "inputs": [ + "cat_13", + "cache_update_2", + "cache_update_3", + "attention_mask" + ], + "output": "matmul_4", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f5b2ba528c9d245492ad94b337e8cafcc65deeb4a5e7975a55ca79cbfe0a29d4.wgsl" + }, + { + "node": "transpose_40", + "inputs": [ + "matmul_4" + ], + "output": "transpose_40", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2792ae5841828bd5af938c10096abe385e63bb3ec2a22e850f146739433c12bf.wgsl" + }, + { + "node": "sigmoid_7", + "inputs": [ + "slice_69" + ], + "output": "sigmoid_7", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a53c1ce51ebaec593ad667efae9603efe3807db8f0ee060a620ac031cc4f2b92.wgsl" + }, + { + "node": "mul_106", + "inputs": [ + "transpose_40", + "sigmoid_7" + ], + "output": "mul_106", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/45a1fd3ed514f208718651026bf2ee30dda66b02cdd950dc71e88eb6bdac0b10.wgsl" + }, + { + "node": "linear_58", + "inputs": [ + "mul_106", + "p_model_model_language_model_layers_7_self_attn_o_proj_weight", + "p_model_model_language_model_layers_7_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_58", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_82", + "inputs": [ + "add_72", + "linear_58" + ], + "output": "add_82", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_108", + "inputs": [ + "add_82", + "p_model_model_language_model_layers_7_post_attention_layernorm_weight" + ], + "output": "mul_108", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_59", + "inputs": [ + "mul_108", + "p_model_model_language_model_layers_7_mlp_gate_proj_weight", + "p_model_model_language_model_layers_7_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_59", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_60", + "inputs": [ + "mul_108", + "p_model_model_language_model_layers_7_mlp_up_proj_weight", + "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_60", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_109", + "inputs": [ + "linear_59", + "linear_60" + ], + "output": "mul_109", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_61", + "inputs": [ + "mul_109", + "p_model_model_language_model_layers_7_mlp_down_proj_weight", + "p_model_model_language_model_layers_7_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_61", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_85", + "inputs": [ + "add_82", + "linear_61" + ], + "output": "add_85", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_111", + "inputs": [ + "add_85", + "p_model_model_language_model_layers_8_input_layernorm_weight" + ], + "output": "mul_111", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_62", + "inputs": [ + "mul_111", + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_62", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_41", + "inputs": [ + "linear_62" + ], + "output": "transpose_41", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_63", + "inputs": [ + "mul_111", + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_63", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_64", + "inputs": [ + "mul_111", + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_64", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_65", + "inputs": [ + "mul_111", + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_65", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_15", + "inputs": [ + "states_16", + "transpose_41" + ], + "output": "cat_15", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_78", + "inputs": [ + "cat_15" + ], + "output": "slice_78", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_16" + }, + { + "node": "conv1d_6", + "inputs": [ + "cat_15", + "p_model_model_language_model_layers_8_linear_attn_conv1d_weight" + ], + "output": "conv1d_6", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_79", + "inputs": [ + "conv1d_6" + ], + "output": "slice_79", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_20", + "inputs": [ + "slice_79" + ], + "output": "silu_20", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_80", + "inputs": [ + "silu_20" + ], + "output": "slice_80", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_42", + "inputs": [ + "slice_80" + ], + "output": "transpose_42", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_81", + "inputs": [ + "transpose_42" + ], + "output": "slice_81", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_82", + "inputs": [ + "transpose_42" + ], + "output": "slice_82", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_83", + "inputs": [ + "transpose_42" + ], + "output": "slice_83", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_8", + "inputs": [ + "linear_64" + ], + "output": "sigmoid_8", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_6", + "inputs": [ + "p_model_model_language_model_layers_8_linear_attn_a_log" + ], + "output": "exp_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_10", + "inputs": [ + "exp_6" + ], + "output": "neg_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_88", + "inputs": [ + "linear_65", + "p_model_model_language_model_layers_8_linear_attn_dt_bias" + ], + "output": "add_88", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_6", + "inputs": [ + "add_88" + ], + "output": "softplus_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_112", + "inputs": [ + "neg_10", + "softplus_6" + ], + "output": "mul_112", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_113", + "inputs": [ + "slice_81" + ], + "output": "mul_113", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_13", + "inputs": [ + "mul_113" + ], + "output": "sum_13", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_89", + "inputs": [ + "sum_13" + ], + "output": "add_89", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_39", + "inputs": [ + "add_89" + ], + "output": "rsqrt_39", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_114", + "inputs": [ + "slice_81", + "rsqrt_39" + ], + "output": "mul_114", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_115", + "inputs": [ + "slice_82" + ], + "output": "mul_115", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_14", + "inputs": [ + "mul_115" + ], + "output": "sum_14", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_90", + "inputs": [ + "sum_14" + ], + "output": "add_90", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_40", + "inputs": [ + "add_90" + ], + "output": "rsqrt_40", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_116", + "inputs": [ + "slice_82", + "rsqrt_40" + ], + "output": "mul_116", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_6", + "inputs": [ + "mul_114", + "mul_116", + "slice_83", + "mul_112", + "sigmoid_8", + "states_17" + ], + "output": "normalized_gated_delta_rule_6", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_84", + "inputs": [ + "normalized_gated_delta_rule_6" + ], + "output": "slice_84", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_17" + }, + { + "node": "slice_85", + "inputs": [ + "normalized_gated_delta_rule_6" + ], + "output": "slice_85", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_43", + "inputs": [ + "slice_85" + ], + "output": "transpose_43", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_118", + "inputs": [ + "transpose_43", + "p_model_model_language_model_layers_8_linear_attn_norm_weight" + ], + "output": "mul_118", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_119", + "inputs": [ + "linear_63", + "mul_118" + ], + "output": "mul_119", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_66", + "inputs": [ + "mul_119", + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_66", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_92", + "inputs": [ + "add_85", + "linear_66" + ], + "output": "add_92", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_121", + "inputs": [ + "add_92", + "p_model_model_language_model_layers_8_post_attention_layernorm_weight" + ], + "output": "mul_121", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_67", + "inputs": [ + "mul_121", + "p_model_model_language_model_layers_8_mlp_gate_proj_weight", + "p_model_model_language_model_layers_8_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_67", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_68", + "inputs": [ + "mul_121", + "p_model_model_language_model_layers_8_mlp_up_proj_weight", + "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_68", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_122", + "inputs": [ + "linear_67", + "linear_68" + ], + "output": "mul_122", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_69", + "inputs": [ + "mul_122", + "p_model_model_language_model_layers_8_mlp_down_proj_weight", + "p_model_model_language_model_layers_8_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_69", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_95", + "inputs": [ + "add_92", + "linear_69" + ], + "output": "add_95", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_124", + "inputs": [ + "add_95", + "p_model_model_language_model_layers_9_input_layernorm_weight" + ], + "output": "mul_124", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_70", + "inputs": [ + "mul_124", + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_70", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_44", + "inputs": [ + "linear_70" + ], + "output": "transpose_44", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_71", + "inputs": [ + "mul_124", + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_71", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_72", + "inputs": [ + "mul_124", + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_72", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_73", + "inputs": [ + "mul_124", + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_73", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_16", + "inputs": [ + "states_18", + "transpose_44" + ], + "output": "cat_16", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_86", + "inputs": [ + "cat_16" + ], + "output": "slice_86", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_18" + }, + { + "node": "conv1d_7", + "inputs": [ + "cat_16", + "p_model_model_language_model_layers_9_linear_attn_conv1d_weight" + ], + "output": "conv1d_7", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_87", + "inputs": [ + "conv1d_7" + ], + "output": "slice_87", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_23", + "inputs": [ + "slice_87" + ], + "output": "silu_23", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_88", + "inputs": [ + "silu_23" + ], + "output": "slice_88", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_45", + "inputs": [ + "slice_88" + ], + "output": "transpose_45", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_89", + "inputs": [ + "transpose_45" + ], + "output": "slice_89", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_90", + "inputs": [ + "transpose_45" + ], + "output": "slice_90", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_91", + "inputs": [ + "transpose_45" + ], + "output": "slice_91", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_9", + "inputs": [ + "linear_72" + ], + "output": "sigmoid_9", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_7", + "inputs": [ + "p_model_model_language_model_layers_9_linear_attn_a_log" + ], + "output": "exp_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_11", + "inputs": [ + "exp_7" + ], + "output": "neg_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_98", + "inputs": [ + "linear_73", + "p_model_model_language_model_layers_9_linear_attn_dt_bias" + ], + "output": "add_98", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_7", + "inputs": [ + "add_98" + ], + "output": "softplus_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_125", + "inputs": [ + "neg_11", + "softplus_7" + ], + "output": "mul_125", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_126", + "inputs": [ + "slice_89" + ], + "output": "mul_126", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_15", + "inputs": [ + "mul_126" + ], + "output": "sum_15", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_99", + "inputs": [ + "sum_15" + ], + "output": "add_99", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_44", + "inputs": [ + "add_99" + ], + "output": "rsqrt_44", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_127", + "inputs": [ + "slice_89", + "rsqrt_44" + ], + "output": "mul_127", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_128", + "inputs": [ + "slice_90" + ], + "output": "mul_128", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_16", + "inputs": [ + "mul_128" + ], + "output": "sum_16", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_100", + "inputs": [ + "sum_16" + ], + "output": "add_100", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_45", + "inputs": [ + "add_100" + ], + "output": "rsqrt_45", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_129", + "inputs": [ + "slice_90", + "rsqrt_45" + ], + "output": "mul_129", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_7", + "inputs": [ + "mul_127", + "mul_129", + "slice_91", + "mul_125", + "sigmoid_9", + "states_19" + ], + "output": "normalized_gated_delta_rule_7", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_92", + "inputs": [ + "normalized_gated_delta_rule_7" + ], + "output": "slice_92", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_19" + }, + { + "node": "slice_93", + "inputs": [ + "normalized_gated_delta_rule_7" + ], + "output": "slice_93", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_46", + "inputs": [ + "slice_93" + ], + "output": "transpose_46", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_131", + "inputs": [ + "transpose_46", + "p_model_model_language_model_layers_9_linear_attn_norm_weight" + ], + "output": "mul_131", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_132", + "inputs": [ + "linear_71", + "mul_131" + ], + "output": "mul_132", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_74", + "inputs": [ + "mul_132", + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_74", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_102", + "inputs": [ + "add_95", + "linear_74" + ], + "output": "add_102", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_134", + "inputs": [ + "add_102", + "p_model_model_language_model_layers_9_post_attention_layernorm_weight" + ], + "output": "mul_134", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_75", + "inputs": [ + "mul_134", + "p_model_model_language_model_layers_9_mlp_gate_proj_weight", + "p_model_model_language_model_layers_9_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_75", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_76", + "inputs": [ + "mul_134", + "p_model_model_language_model_layers_9_mlp_up_proj_weight", + "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_76", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_135", + "inputs": [ + "linear_75", + "linear_76" + ], + "output": "mul_135", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_77", + "inputs": [ + "mul_135", + "p_model_model_language_model_layers_9_mlp_down_proj_weight", + "p_model_model_language_model_layers_9_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_77", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_105", + "inputs": [ + "add_102", + "linear_77" + ], + "output": "add_105", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_137", + "inputs": [ + "add_105", + "p_model_model_language_model_layers_10_input_layernorm_weight" + ], + "output": "mul_137", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_78", + "inputs": [ + "mul_137", + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_78", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_47", + "inputs": [ + "linear_78" + ], + "output": "transpose_47", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_79", + "inputs": [ + "mul_137", + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_79", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_80", + "inputs": [ + "mul_137", + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_80", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_81", + "inputs": [ + "mul_137", + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_81", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_17", + "inputs": [ + "states_20", + "transpose_47" + ], + "output": "cat_17", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_94", + "inputs": [ + "cat_17" + ], + "output": "slice_94", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_20" + }, + { + "node": "conv1d_8", + "inputs": [ + "cat_17", + "p_model_model_language_model_layers_10_linear_attn_conv1d_weight" + ], + "output": "conv1d_8", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_95", + "inputs": [ + "conv1d_8" + ], + "output": "slice_95", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_26", + "inputs": [ + "slice_95" + ], + "output": "silu_26", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_96", + "inputs": [ + "silu_26" + ], + "output": "slice_96", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_48", + "inputs": [ + "slice_96" + ], + "output": "transpose_48", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_97", + "inputs": [ + "transpose_48" + ], + "output": "slice_97", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_98", + "inputs": [ + "transpose_48" + ], + "output": "slice_98", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_99", + "inputs": [ + "transpose_48" + ], + "output": "slice_99", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_10", + "inputs": [ + "linear_80" + ], + "output": "sigmoid_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_8", + "inputs": [ + "p_model_model_language_model_layers_10_linear_attn_a_log" + ], + "output": "exp_8", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_12", + "inputs": [ + "exp_8" + ], + "output": "neg_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_108", + "inputs": [ + "linear_81", + "p_model_model_language_model_layers_10_linear_attn_dt_bias" + ], + "output": "add_108", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_8", + "inputs": [ + "add_108" + ], + "output": "softplus_8", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_138", + "inputs": [ + "neg_12", + "softplus_8" + ], + "output": "mul_138", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_139", + "inputs": [ + "slice_97" + ], + "output": "mul_139", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_17", + "inputs": [ + "mul_139" + ], + "output": "sum_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_109", + "inputs": [ + "sum_17" + ], + "output": "add_109", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_49", + "inputs": [ + "add_109" + ], + "output": "rsqrt_49", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_140", + "inputs": [ + "slice_97", + "rsqrt_49" + ], + "output": "mul_140", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_141", + "inputs": [ + "slice_98" + ], + "output": "mul_141", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_18", + "inputs": [ + "mul_141" + ], + "output": "sum_18", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_110", + "inputs": [ + "sum_18" + ], + "output": "add_110", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_50", + "inputs": [ + "add_110" + ], + "output": "rsqrt_50", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_142", + "inputs": [ + "slice_98", + "rsqrt_50" + ], + "output": "mul_142", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_8", + "inputs": [ + "mul_140", + "mul_142", + "slice_99", + "mul_138", + "sigmoid_10", + "states_21" + ], + "output": "normalized_gated_delta_rule_8", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_100", + "inputs": [ + "normalized_gated_delta_rule_8" + ], + "output": "slice_100", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_21" + }, + { + "node": "slice_101", + "inputs": [ + "normalized_gated_delta_rule_8" + ], + "output": "slice_101", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_49", + "inputs": [ + "slice_101" + ], + "output": "transpose_49", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_144", + "inputs": [ + "transpose_49", + "p_model_model_language_model_layers_10_linear_attn_norm_weight" + ], + "output": "mul_144", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_145", + "inputs": [ + "linear_79", + "mul_144" + ], + "output": "mul_145", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_82", + "inputs": [ + "mul_145", + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_82", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_112", + "inputs": [ + "add_105", + "linear_82" + ], + "output": "add_112", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_147", + "inputs": [ + "add_112", + "p_model_model_language_model_layers_10_post_attention_layernorm_weight" + ], + "output": "mul_147", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_83", + "inputs": [ + "mul_147", + "p_model_model_language_model_layers_10_mlp_gate_proj_weight", + "p_model_model_language_model_layers_10_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_83", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_84", + "inputs": [ + "mul_147", + "p_model_model_language_model_layers_10_mlp_up_proj_weight", + "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_84", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_148", + "inputs": [ + "linear_83", + "linear_84" + ], + "output": "mul_148", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_85", + "inputs": [ + "mul_148", + "p_model_model_language_model_layers_10_mlp_down_proj_weight", + "p_model_model_language_model_layers_10_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_85", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_115", + "inputs": [ + "add_112", + "linear_85" + ], + "output": "add_115", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_150", + "inputs": [ + "add_115", + "p_model_model_language_model_layers_11_input_layernorm_weight" + ], + "output": "mul_150", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_86", + "inputs": [ + "mul_150", + "p_model_model_language_model_layers_11_self_attn_q_proj_weight", + "p_model_model_language_model_layers_11_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_86", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/f7f0ee18ecdc712d36e487eef201cb442f28f824e5b6fac7e4d16cb6b7debec4.wgsl" + }, + { + "node": "slice_102", + "inputs": [ + "linear_86" + ], + "output": "slice_102", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2eeb54034621ff097879064a60fd4e0dd5d68201e48296ad234413a39579acc8.wgsl" + }, + { + "node": "slice_103", + "inputs": [ + "linear_86" + ], + "output": "slice_103", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0d1edb63bf4dc212d55f8a4c79a50e04c0d822bd9f6e193003e2071ba0d57d69.wgsl" + }, + { + "node": "mul_152", + "inputs": [ + "slice_102", + "p_model_model_language_model_layers_11_self_attn_q_norm_weight" + ], + "output": "mul_152", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl" + }, + { + "node": "transpose_50", + "inputs": [ + "mul_152" + ], + "output": "transpose_50", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/18a0b02a12ebf73203283d267342582d54ba411aeee9cbe21ad89ae890e4cfea.wgsl" + }, + { + "node": "linear_87", + "inputs": [ + "mul_150", + "p_model_model_language_model_layers_11_self_attn_k_proj_weight", + "p_model_model_language_model_layers_11_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_87", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl" + }, + { + "node": "mul_154", + "inputs": [ + "linear_87", + "p_model_model_language_model_layers_11_self_attn_k_norm_weight" + ], + "output": "mul_154", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/f994142054bcb47d6a554524dd8de9c3054467a1d07011027c47f475d0909c19.wgsl" + }, + { + "node": "transpose_51", + "inputs": [ + "mul_154" + ], + "output": "transpose_51", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl" + }, + { + "node": "linear_88", + "inputs": [ + "mul_150", + "p_model_model_language_model_layers_11_self_attn_v_proj_weight", + "p_model_model_language_model_layers_11_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_88", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl" + }, + { + "node": "transpose_52", + "inputs": [ + "linear_88" + ], + "output": "transpose_52", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl" + }, + { + "node": "slice_104", + "inputs": [ + "transpose_50" + ], + "output": "slice_104", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8b5e4c9b7889f2bccebf60987b6f5b8c23542c3475cd8d78656dc7ecffd81f8d.wgsl" + }, + { + "node": "slice_105", + "inputs": [ + "transpose_50" + ], + "output": "slice_105", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/f95724b4a3d2dbafaccd23a45f68234bb49e4a39168942060ca628960993f33d.wgsl" + }, + { + "node": "slice_106", + "inputs": [ + "transpose_51" + ], + "output": "slice_106", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/ad72f60d58133272228319f1c066e7cd43ac28573ddeeb7a6f69be01af341f57.wgsl" + }, + { + "node": "slice_107", + "inputs": [ + "transpose_51" + ], + "output": "slice_107", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/8af32f59dfb78c79ac2187ae707f0897b89fb8bb0e3d832e2edac308eefa603c.wgsl" + }, + { + "node": "mul_155", + "inputs": [ + "slice_104", + "mul" + ], + "output": "mul_155", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl" + }, + { + "node": "slice_108", + "inputs": [ + "slice_104" + ], + "output": "slice_108", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8ee286e06c1271162bea74e14b4bd5b1d66890648217c79435e6b3f1d4181ec8.wgsl" + }, + { + "node": "slice_109", + "inputs": [ + "slice_104" + ], + "output": "slice_109", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/80d22263b082805bc9d4499247f92b7a1a783a3e1cdd9c652d304492a1cfd425.wgsl" + }, + { + "node": "neg_13", + "inputs": [ + "slice_109" + ], + "output": "neg_13", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl" + }, + { + "node": "cat_18", + "inputs": [ + "neg_13", + "slice_108" + ], + "output": "cat_18", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/bda7ebd3bcf25c31938fca1a582e82b972ab742f85f826f085417b0e6f651e60.wgsl" + }, + { + "node": "mul_156", + "inputs": [ + "cat_18", + "mul_1" + ], + "output": "mul_156", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl" + }, + { + "node": "add_122", + "inputs": [ + "mul_155", + "mul_156" + ], + "output": "add_122", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl" + }, + { + "node": "mul_157", + "inputs": [ + "slice_106", + "mul" + ], + "output": "mul_157", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/d0be84d22b9e3b0cf73729e9a6fad1c6cf7d03fa01b45b8765ed70158578f2a9.wgsl" + }, + { + "node": "slice_110", + "inputs": [ + "slice_106" + ], + "output": "slice_110", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8c3a77238461b0ae13318bad504e1e0779f3f47d1782fe0ca78f20112c031146.wgsl" + }, + { + "node": "slice_111", + "inputs": [ + "slice_106" + ], + "output": "slice_111", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/b2ae1b161c8a6315f8863e47511607027dfe583aac1db4c1626371ddfa0c64e2.wgsl" + }, + { + "node": "neg_14", + "inputs": [ + "slice_111" + ], + "output": "neg_14", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/4fad29e3e337840e25132f10b3e3d55225f727ae13525ff0b0bdf31e03f09d74.wgsl" + }, + { + "node": "cat_19", + "inputs": [ + "neg_14", + "slice_110" + ], + "output": "cat_19", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/898086c99e03b2cdb78e3e5658c3fff202f1c87d09d8da08659bd1ba85e01b09.wgsl" + }, + { + "node": "mul_158", + "inputs": [ + "cat_19", + "mul_1" + ], + "output": "mul_158", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/d0be84d22b9e3b0cf73729e9a6fad1c6cf7d03fa01b45b8765ed70158578f2a9.wgsl" + }, + { + "node": "add_123", + "inputs": [ + "mul_157", + "mul_158" + ], + "output": "add_123", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/f5b51b1e0e52bd87a70a0ca169186ebb5734cbdc2b8f375ba0241485b1df4e66.wgsl" + }, + { + "node": "cat_20", + "inputs": [ + "add_122", + "slice_105" + ], + "output": "cat_20", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl" + }, + { + "node": "cat_21", + "inputs": [ + "add_123", + "slice_107" + ], + "output": "cat_21", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f276cf6656b999339df464b61934487f5c8eb7a0957cf6d902254ab1aa2ffcd0.wgsl" + }, + { + "node": "cache_update_4", + "inputs": [ + "cat_21", + "position_ids" + ], + "output": "cache_update_4", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_22", + "kernel": "kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl" + }, + { + "node": "cache_update_5", + "inputs": [ + "transpose_52", + "position_ids" + ], + "output": "cache_update_5", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_23", + "kernel": "kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl" + }, + { + "node": "matmul_6", + "inputs": [ + "cat_20", + "cache_update_4", + "cache_update_5", + "attention_mask" + ], + "output": "matmul_6", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f5b2ba528c9d245492ad94b337e8cafcc65deeb4a5e7975a55ca79cbfe0a29d4.wgsl" + }, + { + "node": "transpose_54", + "inputs": [ + "matmul_6" + ], + "output": "transpose_54", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2792ae5841828bd5af938c10096abe385e63bb3ec2a22e850f146739433c12bf.wgsl" + }, + { + "node": "sigmoid_11", + "inputs": [ + "slice_103" + ], + "output": "sigmoid_11", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a53c1ce51ebaec593ad667efae9603efe3807db8f0ee060a620ac031cc4f2b92.wgsl" + }, + { + "node": "mul_160", + "inputs": [ + "transpose_54", + "sigmoid_11" + ], + "output": "mul_160", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/45a1fd3ed514f208718651026bf2ee30dda66b02cdd950dc71e88eb6bdac0b10.wgsl" + }, + { + "node": "linear_89", + "inputs": [ + "mul_160", + "p_model_model_language_model_layers_11_self_attn_o_proj_weight", + "p_model_model_language_model_layers_11_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_89", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_125", + "inputs": [ + "add_115", + "linear_89" + ], + "output": "add_125", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_162", + "inputs": [ + "add_125", + "p_model_model_language_model_layers_11_post_attention_layernorm_weight" + ], + "output": "mul_162", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_90", + "inputs": [ + "mul_162", + "p_model_model_language_model_layers_11_mlp_gate_proj_weight", + "p_model_model_language_model_layers_11_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_90", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_91", + "inputs": [ + "mul_162", + "p_model_model_language_model_layers_11_mlp_up_proj_weight", + "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_91", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_163", + "inputs": [ + "linear_90", + "linear_91" + ], + "output": "mul_163", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_92", + "inputs": [ + "mul_163", + "p_model_model_language_model_layers_11_mlp_down_proj_weight", + "p_model_model_language_model_layers_11_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_92", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_128", + "inputs": [ + "add_125", + "linear_92" + ], + "output": "add_128", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_165", + "inputs": [ + "add_128", + "p_model_model_language_model_layers_12_input_layernorm_weight" + ], + "output": "mul_165", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_93", + "inputs": [ + "mul_165", + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_93", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_55", + "inputs": [ + "linear_93" + ], + "output": "transpose_55", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_94", + "inputs": [ + "mul_165", + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_94", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_95", + "inputs": [ + "mul_165", + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_95", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_96", + "inputs": [ + "mul_165", + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_96", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_22", + "inputs": [ + "states_24", + "transpose_55" + ], + "output": "cat_22", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_112", + "inputs": [ + "cat_22" + ], + "output": "slice_112", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_24" + }, + { + "node": "conv1d_9", + "inputs": [ + "cat_22", + "p_model_model_language_model_layers_12_linear_attn_conv1d_weight" + ], + "output": "conv1d_9", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_113", + "inputs": [ + "conv1d_9" + ], + "output": "slice_113", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_30", + "inputs": [ + "slice_113" + ], + "output": "silu_30", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_114", + "inputs": [ + "silu_30" + ], + "output": "slice_114", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_56", + "inputs": [ + "slice_114" + ], + "output": "transpose_56", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_115", + "inputs": [ + "transpose_56" + ], + "output": "slice_115", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_116", + "inputs": [ + "transpose_56" + ], + "output": "slice_116", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_117", + "inputs": [ + "transpose_56" + ], + "output": "slice_117", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_12", + "inputs": [ + "linear_95" + ], + "output": "sigmoid_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_9", + "inputs": [ + "p_model_model_language_model_layers_12_linear_attn_a_log" + ], + "output": "exp_9", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_15", + "inputs": [ + "exp_9" + ], + "output": "neg_15", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_131", + "inputs": [ + "linear_96", + "p_model_model_language_model_layers_12_linear_attn_dt_bias" + ], + "output": "add_131", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_9", + "inputs": [ + "add_131" + ], + "output": "softplus_9", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_166", + "inputs": [ + "neg_15", + "softplus_9" + ], + "output": "mul_166", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_167", + "inputs": [ + "slice_115" + ], + "output": "mul_167", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_19", + "inputs": [ + "mul_167" + ], + "output": "sum_19", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_132", + "inputs": [ + "sum_19" + ], + "output": "add_132", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_58", + "inputs": [ + "add_132" + ], + "output": "rsqrt_58", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_168", + "inputs": [ + "slice_115", + "rsqrt_58" + ], + "output": "mul_168", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_169", + "inputs": [ + "slice_116" + ], + "output": "mul_169", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_20", + "inputs": [ + "mul_169" + ], + "output": "sum_20", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_133", + "inputs": [ + "sum_20" + ], + "output": "add_133", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_59", + "inputs": [ + "add_133" + ], + "output": "rsqrt_59", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_170", + "inputs": [ + "slice_116", + "rsqrt_59" + ], + "output": "mul_170", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_9", + "inputs": [ + "mul_168", + "mul_170", + "slice_117", + "mul_166", + "sigmoid_12", + "states_25" + ], + "output": "normalized_gated_delta_rule_9", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_118", + "inputs": [ + "normalized_gated_delta_rule_9" + ], + "output": "slice_118", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_25" + }, + { + "node": "slice_119", + "inputs": [ + "normalized_gated_delta_rule_9" + ], + "output": "slice_119", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_57", + "inputs": [ + "slice_119" + ], + "output": "transpose_57", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_172", + "inputs": [ + "transpose_57", + "p_model_model_language_model_layers_12_linear_attn_norm_weight" + ], + "output": "mul_172", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_173", + "inputs": [ + "linear_94", + "mul_172" + ], + "output": "mul_173", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_97", + "inputs": [ + "mul_173", + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_97", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_135", + "inputs": [ + "add_128", + "linear_97" + ], + "output": "add_135", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_175", + "inputs": [ + "add_135", + "p_model_model_language_model_layers_12_post_attention_layernorm_weight" + ], + "output": "mul_175", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_98", + "inputs": [ + "mul_175", + "p_model_model_language_model_layers_12_mlp_gate_proj_weight", + "p_model_model_language_model_layers_12_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_98", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_99", + "inputs": [ + "mul_175", + "p_model_model_language_model_layers_12_mlp_up_proj_weight", + "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_99", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_176", + "inputs": [ + "linear_98", + "linear_99" + ], + "output": "mul_176", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_100", + "inputs": [ + "mul_176", + "p_model_model_language_model_layers_12_mlp_down_proj_weight", + "p_model_model_language_model_layers_12_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_100", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_138", + "inputs": [ + "add_135", + "linear_100" + ], + "output": "add_138", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_178", + "inputs": [ + "add_138", + "p_model_model_language_model_layers_13_input_layernorm_weight" + ], + "output": "mul_178", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_101", + "inputs": [ + "mul_178", + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_101", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_58", + "inputs": [ + "linear_101" + ], + "output": "transpose_58", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_102", + "inputs": [ + "mul_178", + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_102", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_103", + "inputs": [ + "mul_178", + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_103", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_104", + "inputs": [ + "mul_178", + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_104", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_23", + "inputs": [ + "states_26", + "transpose_58" + ], + "output": "cat_23", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_120", + "inputs": [ + "cat_23" + ], + "output": "slice_120", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_26" + }, + { + "node": "conv1d_10", + "inputs": [ + "cat_23", + "p_model_model_language_model_layers_13_linear_attn_conv1d_weight" + ], + "output": "conv1d_10", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_121", + "inputs": [ + "conv1d_10" + ], + "output": "slice_121", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_33", + "inputs": [ + "slice_121" + ], + "output": "silu_33", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_122", + "inputs": [ + "silu_33" + ], + "output": "slice_122", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_59", + "inputs": [ + "slice_122" + ], + "output": "transpose_59", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_123", + "inputs": [ + "transpose_59" + ], + "output": "slice_123", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_124", + "inputs": [ + "transpose_59" + ], + "output": "slice_124", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_125", + "inputs": [ + "transpose_59" + ], + "output": "slice_125", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_13", + "inputs": [ + "linear_103" + ], + "output": "sigmoid_13", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_10", + "inputs": [ + "p_model_model_language_model_layers_13_linear_attn_a_log" + ], + "output": "exp_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_16", + "inputs": [ + "exp_10" + ], + "output": "neg_16", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_141", + "inputs": [ + "linear_104", + "p_model_model_language_model_layers_13_linear_attn_dt_bias" + ], + "output": "add_141", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_10", + "inputs": [ + "add_141" + ], + "output": "softplus_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_179", + "inputs": [ + "neg_16", + "softplus_10" + ], + "output": "mul_179", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_180", + "inputs": [ + "slice_123" + ], + "output": "mul_180", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_21", + "inputs": [ + "mul_180" + ], + "output": "sum_21", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_142", + "inputs": [ + "sum_21" + ], + "output": "add_142", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_63", + "inputs": [ + "add_142" + ], + "output": "rsqrt_63", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_181", + "inputs": [ + "slice_123", + "rsqrt_63" + ], + "output": "mul_181", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_182", + "inputs": [ + "slice_124" + ], + "output": "mul_182", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_22", + "inputs": [ + "mul_182" + ], + "output": "sum_22", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_143", + "inputs": [ + "sum_22" + ], + "output": "add_143", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_64", + "inputs": [ + "add_143" + ], + "output": "rsqrt_64", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_183", + "inputs": [ + "slice_124", + "rsqrt_64" + ], + "output": "mul_183", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_10", + "inputs": [ + "mul_181", + "mul_183", + "slice_125", + "mul_179", + "sigmoid_13", + "states_27" + ], + "output": "normalized_gated_delta_rule_10", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_126", + "inputs": [ + "normalized_gated_delta_rule_10" + ], + "output": "slice_126", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_27" + }, + { + "node": "slice_127", + "inputs": [ + "normalized_gated_delta_rule_10" + ], + "output": "slice_127", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_60", + "inputs": [ + "slice_127" + ], + "output": "transpose_60", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_185", + "inputs": [ + "transpose_60", + "p_model_model_language_model_layers_13_linear_attn_norm_weight" + ], + "output": "mul_185", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_186", + "inputs": [ + "linear_102", + "mul_185" + ], + "output": "mul_186", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_105", + "inputs": [ + "mul_186", + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_105", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_145", + "inputs": [ + "add_138", + "linear_105" + ], + "output": "add_145", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_188", + "inputs": [ + "add_145", + "p_model_model_language_model_layers_13_post_attention_layernorm_weight" + ], + "output": "mul_188", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_106", + "inputs": [ + "mul_188", + "p_model_model_language_model_layers_13_mlp_gate_proj_weight", + "p_model_model_language_model_layers_13_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_106", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_107", + "inputs": [ + "mul_188", + "p_model_model_language_model_layers_13_mlp_up_proj_weight", + "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_107", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_189", + "inputs": [ + "linear_106", + "linear_107" + ], + "output": "mul_189", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_108", + "inputs": [ + "mul_189", + "p_model_model_language_model_layers_13_mlp_down_proj_weight", + "p_model_model_language_model_layers_13_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_108", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_148", + "inputs": [ + "add_145", + "linear_108" + ], + "output": "add_148", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_191", + "inputs": [ + "add_148", + "p_model_model_language_model_layers_14_input_layernorm_weight" + ], + "output": "mul_191", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_109", + "inputs": [ + "mul_191", + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_109", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_61", + "inputs": [ + "linear_109" + ], + "output": "transpose_61", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_110", + "inputs": [ + "mul_191", + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_110", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_111", + "inputs": [ + "mul_191", + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_111", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_112", + "inputs": [ + "mul_191", + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_112", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_24", + "inputs": [ + "states_28", + "transpose_61" + ], + "output": "cat_24", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_128", + "inputs": [ + "cat_24" + ], + "output": "slice_128", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_28" + }, + { + "node": "conv1d_11", + "inputs": [ + "cat_24", + "p_model_model_language_model_layers_14_linear_attn_conv1d_weight" + ], + "output": "conv1d_11", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_129", + "inputs": [ + "conv1d_11" + ], + "output": "slice_129", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_36", + "inputs": [ + "slice_129" + ], + "output": "silu_36", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_130", + "inputs": [ + "silu_36" + ], + "output": "slice_130", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_62", + "inputs": [ + "slice_130" + ], + "output": "transpose_62", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_131", + "inputs": [ + "transpose_62" + ], + "output": "slice_131", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_132", + "inputs": [ + "transpose_62" + ], + "output": "slice_132", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_133", + "inputs": [ + "transpose_62" + ], + "output": "slice_133", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_14", + "inputs": [ + "linear_111" + ], + "output": "sigmoid_14", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_11", + "inputs": [ + "p_model_model_language_model_layers_14_linear_attn_a_log" + ], + "output": "exp_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_17", + "inputs": [ + "exp_11" + ], + "output": "neg_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_151", + "inputs": [ + "linear_112", + "p_model_model_language_model_layers_14_linear_attn_dt_bias" + ], + "output": "add_151", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_11", + "inputs": [ + "add_151" + ], + "output": "softplus_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_192", + "inputs": [ + "neg_17", + "softplus_11" + ], + "output": "mul_192", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_193", + "inputs": [ + "slice_131" + ], + "output": "mul_193", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_23", + "inputs": [ + "mul_193" + ], + "output": "sum_23", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_152", + "inputs": [ + "sum_23" + ], + "output": "add_152", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_68", + "inputs": [ + "add_152" + ], + "output": "rsqrt_68", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_194", + "inputs": [ + "slice_131", + "rsqrt_68" + ], + "output": "mul_194", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_195", + "inputs": [ + "slice_132" + ], + "output": "mul_195", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_24", + "inputs": [ + "mul_195" + ], + "output": "sum_24", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_153", + "inputs": [ + "sum_24" + ], + "output": "add_153", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_69", + "inputs": [ + "add_153" + ], + "output": "rsqrt_69", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_196", + "inputs": [ + "slice_132", + "rsqrt_69" + ], + "output": "mul_196", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_11", + "inputs": [ + "mul_194", + "mul_196", + "slice_133", + "mul_192", + "sigmoid_14", + "states_29" + ], + "output": "normalized_gated_delta_rule_11", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_134", + "inputs": [ + "normalized_gated_delta_rule_11" + ], + "output": "slice_134", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_29" + }, + { + "node": "slice_135", + "inputs": [ + "normalized_gated_delta_rule_11" + ], + "output": "slice_135", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_63", + "inputs": [ + "slice_135" + ], + "output": "transpose_63", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_198", + "inputs": [ + "transpose_63", + "p_model_model_language_model_layers_14_linear_attn_norm_weight" + ], + "output": "mul_198", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_199", + "inputs": [ + "linear_110", + "mul_198" + ], + "output": "mul_199", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_113", + "inputs": [ + "mul_199", + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_113", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_155", + "inputs": [ + "add_148", + "linear_113" + ], + "output": "add_155", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_201", + "inputs": [ + "add_155", + "p_model_model_language_model_layers_14_post_attention_layernorm_weight" + ], + "output": "mul_201", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_114", + "inputs": [ + "mul_201", + "p_model_model_language_model_layers_14_mlp_gate_proj_weight", + "p_model_model_language_model_layers_14_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_114", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_115", + "inputs": [ + "mul_201", + "p_model_model_language_model_layers_14_mlp_up_proj_weight", + "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_115", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_202", + "inputs": [ + "linear_114", + "linear_115" + ], + "output": "mul_202", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_116", + "inputs": [ + "mul_202", + "p_model_model_language_model_layers_14_mlp_down_proj_weight", + "p_model_model_language_model_layers_14_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_116", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_158", + "inputs": [ + "add_155", + "linear_116" + ], + "output": "add_158", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_204", + "inputs": [ + "add_158", + "p_model_model_language_model_layers_15_input_layernorm_weight" + ], + "output": "mul_204", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_117", + "inputs": [ + "mul_204", + "p_model_model_language_model_layers_15_self_attn_q_proj_weight", + "p_model_model_language_model_layers_15_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_117", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/f7f0ee18ecdc712d36e487eef201cb442f28f824e5b6fac7e4d16cb6b7debec4.wgsl" + }, + { + "node": "slice_136", + "inputs": [ + "linear_117" + ], + "output": "slice_136", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2eeb54034621ff097879064a60fd4e0dd5d68201e48296ad234413a39579acc8.wgsl" + }, + { + "node": "slice_137", + "inputs": [ + "linear_117" + ], + "output": "slice_137", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0d1edb63bf4dc212d55f8a4c79a50e04c0d822bd9f6e193003e2071ba0d57d69.wgsl" + }, + { + "node": "mul_206", + "inputs": [ + "slice_136", + "p_model_model_language_model_layers_15_self_attn_q_norm_weight" + ], + "output": "mul_206", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl" + }, + { + "node": "transpose_64", + "inputs": [ + "mul_206" + ], + "output": "transpose_64", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/18a0b02a12ebf73203283d267342582d54ba411aeee9cbe21ad89ae890e4cfea.wgsl" + }, + { + "node": "linear_118", + "inputs": [ + "mul_204", + "p_model_model_language_model_layers_15_self_attn_k_proj_weight", + "p_model_model_language_model_layers_15_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_118", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl" + }, + { + "node": "mul_208", + "inputs": [ + "linear_118", + "p_model_model_language_model_layers_15_self_attn_k_norm_weight" + ], + "output": "mul_208", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/f994142054bcb47d6a554524dd8de9c3054467a1d07011027c47f475d0909c19.wgsl" + }, + { + "node": "transpose_65", + "inputs": [ + "mul_208" + ], + "output": "transpose_65", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl" + }, + { + "node": "linear_119", + "inputs": [ + "mul_204", + "p_model_model_language_model_layers_15_self_attn_v_proj_weight", + "p_model_model_language_model_layers_15_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_119", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl" + }, + { + "node": "transpose_66", + "inputs": [ + "linear_119" + ], + "output": "transpose_66", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl" + }, + { + "node": "slice_138", + "inputs": [ + "transpose_64" + ], + "output": "slice_138", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8b5e4c9b7889f2bccebf60987b6f5b8c23542c3475cd8d78656dc7ecffd81f8d.wgsl" + }, + { + "node": "slice_139", + "inputs": [ + "transpose_64" + ], + "output": "slice_139", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/f95724b4a3d2dbafaccd23a45f68234bb49e4a39168942060ca628960993f33d.wgsl" + }, + { + "node": "slice_140", + "inputs": [ + "transpose_65" + ], + "output": "slice_140", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/ad72f60d58133272228319f1c066e7cd43ac28573ddeeb7a6f69be01af341f57.wgsl" + }, + { + "node": "slice_141", + "inputs": [ + "transpose_65" + ], + "output": "slice_141", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/8af32f59dfb78c79ac2187ae707f0897b89fb8bb0e3d832e2edac308eefa603c.wgsl" + }, + { + "node": "mul_209", + "inputs": [ + "slice_138", + "mul" + ], + "output": "mul_209", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl" + }, + { + "node": "slice_142", + "inputs": [ + "slice_138" + ], + "output": "slice_142", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8ee286e06c1271162bea74e14b4bd5b1d66890648217c79435e6b3f1d4181ec8.wgsl" + }, + { + "node": "slice_143", + "inputs": [ + "slice_138" + ], + "output": "slice_143", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/80d22263b082805bc9d4499247f92b7a1a783a3e1cdd9c652d304492a1cfd425.wgsl" + }, + { + "node": "neg_18", + "inputs": [ + "slice_143" + ], + "output": "neg_18", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl" + }, + { + "node": "cat_25", + "inputs": [ + "neg_18", + "slice_142" + ], + "output": "cat_25", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/bda7ebd3bcf25c31938fca1a582e82b972ab742f85f826f085417b0e6f651e60.wgsl" + }, + { + "node": "mul_210", + "inputs": [ + "cat_25", + "mul_1" + ], + "output": "mul_210", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl" + }, + { + "node": "add_165", + "inputs": [ + "mul_209", + "mul_210" + ], + "output": "add_165", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl" + }, + { + "node": "mul_211", + "inputs": [ + "slice_140", + "mul" + ], + "output": "mul_211", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/d0be84d22b9e3b0cf73729e9a6fad1c6cf7d03fa01b45b8765ed70158578f2a9.wgsl" + }, + { + "node": "slice_144", + "inputs": [ + "slice_140" + ], + "output": "slice_144", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8c3a77238461b0ae13318bad504e1e0779f3f47d1782fe0ca78f20112c031146.wgsl" + }, + { + "node": "slice_145", + "inputs": [ + "slice_140" + ], + "output": "slice_145", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/b2ae1b161c8a6315f8863e47511607027dfe583aac1db4c1626371ddfa0c64e2.wgsl" + }, + { + "node": "neg_19", + "inputs": [ + "slice_145" + ], + "output": "neg_19", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/4fad29e3e337840e25132f10b3e3d55225f727ae13525ff0b0bdf31e03f09d74.wgsl" + }, + { + "node": "cat_26", + "inputs": [ + "neg_19", + "slice_144" + ], + "output": "cat_26", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/898086c99e03b2cdb78e3e5658c3fff202f1c87d09d8da08659bd1ba85e01b09.wgsl" + }, + { + "node": "mul_212", + "inputs": [ + "cat_26", + "mul_1" + ], + "output": "mul_212", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/d0be84d22b9e3b0cf73729e9a6fad1c6cf7d03fa01b45b8765ed70158578f2a9.wgsl" + }, + { + "node": "add_166", + "inputs": [ + "mul_211", + "mul_212" + ], + "output": "add_166", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/f5b51b1e0e52bd87a70a0ca169186ebb5734cbdc2b8f375ba0241485b1df4e66.wgsl" + }, + { + "node": "cat_27", + "inputs": [ + "add_165", + "slice_139" + ], + "output": "cat_27", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl" + }, + { + "node": "cat_28", + "inputs": [ + "add_166", + "slice_141" + ], + "output": "cat_28", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f276cf6656b999339df464b61934487f5c8eb7a0957cf6d902254ab1aa2ffcd0.wgsl" + }, + { + "node": "cache_update_6", + "inputs": [ + "cat_28", + "position_ids" + ], + "output": "cache_update_6", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_30", + "kernel": "kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl" + }, + { + "node": "cache_update_7", + "inputs": [ + "transpose_66", + "position_ids" + ], + "output": "cache_update_7", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_31", + "kernel": "kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl" + }, + { + "node": "matmul_8", + "inputs": [ + "cat_27", + "cache_update_6", + "cache_update_7", + "attention_mask" + ], + "output": "matmul_8", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f5b2ba528c9d245492ad94b337e8cafcc65deeb4a5e7975a55ca79cbfe0a29d4.wgsl" + }, + { + "node": "transpose_68", + "inputs": [ + "matmul_8" + ], + "output": "transpose_68", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2792ae5841828bd5af938c10096abe385e63bb3ec2a22e850f146739433c12bf.wgsl" + }, + { + "node": "sigmoid_15", + "inputs": [ + "slice_137" + ], + "output": "sigmoid_15", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a53c1ce51ebaec593ad667efae9603efe3807db8f0ee060a620ac031cc4f2b92.wgsl" + }, + { + "node": "mul_214", + "inputs": [ + "transpose_68", + "sigmoid_15" + ], + "output": "mul_214", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/45a1fd3ed514f208718651026bf2ee30dda66b02cdd950dc71e88eb6bdac0b10.wgsl" + }, + { + "node": "linear_120", + "inputs": [ + "mul_214", + "p_model_model_language_model_layers_15_self_attn_o_proj_weight", + "p_model_model_language_model_layers_15_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_120", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_168", + "inputs": [ + "add_158", + "linear_120" + ], + "output": "add_168", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_216", + "inputs": [ + "add_168", + "p_model_model_language_model_layers_15_post_attention_layernorm_weight" + ], + "output": "mul_216", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_121", + "inputs": [ + "mul_216", + "p_model_model_language_model_layers_15_mlp_gate_proj_weight", + "p_model_model_language_model_layers_15_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_121", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_122", + "inputs": [ + "mul_216", + "p_model_model_language_model_layers_15_mlp_up_proj_weight", + "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_122", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_217", + "inputs": [ + "linear_121", + "linear_122" + ], + "output": "mul_217", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_123", + "inputs": [ + "mul_217", + "p_model_model_language_model_layers_15_mlp_down_proj_weight", + "p_model_model_language_model_layers_15_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_123", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_171", + "inputs": [ + "add_168", + "linear_123" + ], + "output": "add_171", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_219", + "inputs": [ + "add_171", + "p_model_model_language_model_layers_16_input_layernorm_weight" + ], + "output": "mul_219", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_124", + "inputs": [ + "mul_219", + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_124", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_69", + "inputs": [ + "linear_124" + ], + "output": "transpose_69", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_125", + "inputs": [ + "mul_219", + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_125", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_126", + "inputs": [ + "mul_219", + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_126", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_127", + "inputs": [ + "mul_219", + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_127", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_29", + "inputs": [ + "states_32", + "transpose_69" + ], + "output": "cat_29", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_146", + "inputs": [ + "cat_29" + ], + "output": "slice_146", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_32" + }, + { + "node": "conv1d_12", + "inputs": [ + "cat_29", + "p_model_model_language_model_layers_16_linear_attn_conv1d_weight" + ], + "output": "conv1d_12", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_147", + "inputs": [ + "conv1d_12" + ], + "output": "slice_147", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_40", + "inputs": [ + "slice_147" + ], + "output": "silu_40", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_148", + "inputs": [ + "silu_40" + ], + "output": "slice_148", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_70", + "inputs": [ + "slice_148" + ], + "output": "transpose_70", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_149", + "inputs": [ + "transpose_70" + ], + "output": "slice_149", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_150", + "inputs": [ + "transpose_70" + ], + "output": "slice_150", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_151", + "inputs": [ + "transpose_70" + ], + "output": "slice_151", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_16", + "inputs": [ + "linear_126" + ], + "output": "sigmoid_16", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_12", + "inputs": [ + "p_model_model_language_model_layers_16_linear_attn_a_log" + ], + "output": "exp_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_20", + "inputs": [ + "exp_12" + ], + "output": "neg_20", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_174", + "inputs": [ + "linear_127", + "p_model_model_language_model_layers_16_linear_attn_dt_bias" + ], + "output": "add_174", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_12", + "inputs": [ + "add_174" + ], + "output": "softplus_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_220", + "inputs": [ + "neg_20", + "softplus_12" + ], + "output": "mul_220", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_221", + "inputs": [ + "slice_149" + ], + "output": "mul_221", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_25", + "inputs": [ + "mul_221" + ], + "output": "sum_25", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_175", + "inputs": [ + "sum_25" + ], + "output": "add_175", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_77", + "inputs": [ + "add_175" + ], + "output": "rsqrt_77", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_222", + "inputs": [ + "slice_149", + "rsqrt_77" + ], + "output": "mul_222", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_223", + "inputs": [ + "slice_150" + ], + "output": "mul_223", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_26", + "inputs": [ + "mul_223" + ], + "output": "sum_26", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_176", + "inputs": [ + "sum_26" + ], + "output": "add_176", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_78", + "inputs": [ + "add_176" + ], + "output": "rsqrt_78", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_224", + "inputs": [ + "slice_150", + "rsqrt_78" + ], + "output": "mul_224", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_12", + "inputs": [ + "mul_222", + "mul_224", + "slice_151", + "mul_220", + "sigmoid_16", + "states_33" + ], + "output": "normalized_gated_delta_rule_12", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_152", + "inputs": [ + "normalized_gated_delta_rule_12" + ], + "output": "slice_152", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_33" + }, + { + "node": "slice_153", + "inputs": [ + "normalized_gated_delta_rule_12" + ], + "output": "slice_153", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_71", + "inputs": [ + "slice_153" + ], + "output": "transpose_71", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_226", + "inputs": [ + "transpose_71", + "p_model_model_language_model_layers_16_linear_attn_norm_weight" + ], + "output": "mul_226", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_227", + "inputs": [ + "linear_125", + "mul_226" + ], + "output": "mul_227", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_128", + "inputs": [ + "mul_227", + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_128", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_178", + "inputs": [ + "add_171", + "linear_128" + ], + "output": "add_178", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_229", + "inputs": [ + "add_178", + "p_model_model_language_model_layers_16_post_attention_layernorm_weight" + ], + "output": "mul_229", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_129", + "inputs": [ + "mul_229", + "p_model_model_language_model_layers_16_mlp_gate_proj_weight", + "p_model_model_language_model_layers_16_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_129", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_130", + "inputs": [ + "mul_229", + "p_model_model_language_model_layers_16_mlp_up_proj_weight", + "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_130", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_230", + "inputs": [ + "linear_129", + "linear_130" + ], + "output": "mul_230", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_131", + "inputs": [ + "mul_230", + "p_model_model_language_model_layers_16_mlp_down_proj_weight", + "p_model_model_language_model_layers_16_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_131", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_181", + "inputs": [ + "add_178", + "linear_131" + ], + "output": "add_181", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_232", + "inputs": [ + "add_181", + "p_model_model_language_model_layers_17_input_layernorm_weight" + ], + "output": "mul_232", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_132", + "inputs": [ + "mul_232", + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_132", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_72", + "inputs": [ + "linear_132" + ], + "output": "transpose_72", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_133", + "inputs": [ + "mul_232", + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_133", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_134", + "inputs": [ + "mul_232", + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_134", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_135", + "inputs": [ + "mul_232", + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_135", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_30", + "inputs": [ + "states_34", + "transpose_72" + ], + "output": "cat_30", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_154", + "inputs": [ + "cat_30" + ], + "output": "slice_154", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_34" + }, + { + "node": "conv1d_13", + "inputs": [ + "cat_30", + "p_model_model_language_model_layers_17_linear_attn_conv1d_weight" + ], + "output": "conv1d_13", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_155", + "inputs": [ + "conv1d_13" + ], + "output": "slice_155", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_43", + "inputs": [ + "slice_155" + ], + "output": "silu_43", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_156", + "inputs": [ + "silu_43" + ], + "output": "slice_156", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_73", + "inputs": [ + "slice_156" + ], + "output": "transpose_73", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_157", + "inputs": [ + "transpose_73" + ], + "output": "slice_157", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_158", + "inputs": [ + "transpose_73" + ], + "output": "slice_158", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_159", + "inputs": [ + "transpose_73" + ], + "output": "slice_159", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_17", + "inputs": [ + "linear_134" + ], + "output": "sigmoid_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_13", + "inputs": [ + "p_model_model_language_model_layers_17_linear_attn_a_log" + ], + "output": "exp_13", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_21", + "inputs": [ + "exp_13" + ], + "output": "neg_21", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_184", + "inputs": [ + "linear_135", + "p_model_model_language_model_layers_17_linear_attn_dt_bias" + ], + "output": "add_184", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_13", + "inputs": [ + "add_184" + ], + "output": "softplus_13", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_233", + "inputs": [ + "neg_21", + "softplus_13" + ], + "output": "mul_233", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_234", + "inputs": [ + "slice_157" + ], + "output": "mul_234", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_27", + "inputs": [ + "mul_234" + ], + "output": "sum_27", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_185", + "inputs": [ + "sum_27" + ], + "output": "add_185", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_82", + "inputs": [ + "add_185" + ], + "output": "rsqrt_82", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_235", + "inputs": [ + "slice_157", + "rsqrt_82" + ], + "output": "mul_235", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_236", + "inputs": [ + "slice_158" + ], + "output": "mul_236", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_28", + "inputs": [ + "mul_236" + ], + "output": "sum_28", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_186", + "inputs": [ + "sum_28" + ], + "output": "add_186", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_83", + "inputs": [ + "add_186" + ], + "output": "rsqrt_83", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_237", + "inputs": [ + "slice_158", + "rsqrt_83" + ], + "output": "mul_237", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_13", + "inputs": [ + "mul_235", + "mul_237", + "slice_159", + "mul_233", + "sigmoid_17", + "states_35" + ], + "output": "normalized_gated_delta_rule_13", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_160", + "inputs": [ + "normalized_gated_delta_rule_13" + ], + "output": "slice_160", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_35" + }, + { + "node": "slice_161", + "inputs": [ + "normalized_gated_delta_rule_13" + ], + "output": "slice_161", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_74", + "inputs": [ + "slice_161" + ], + "output": "transpose_74", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_239", + "inputs": [ + "transpose_74", + "p_model_model_language_model_layers_17_linear_attn_norm_weight" + ], + "output": "mul_239", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_240", + "inputs": [ + "linear_133", + "mul_239" + ], + "output": "mul_240", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_136", + "inputs": [ + "mul_240", + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_136", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_188", + "inputs": [ + "add_181", + "linear_136" + ], + "output": "add_188", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_242", + "inputs": [ + "add_188", + "p_model_model_language_model_layers_17_post_attention_layernorm_weight" + ], + "output": "mul_242", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_137", + "inputs": [ + "mul_242", + "p_model_model_language_model_layers_17_mlp_gate_proj_weight", + "p_model_model_language_model_layers_17_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_137", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_138", + "inputs": [ + "mul_242", + "p_model_model_language_model_layers_17_mlp_up_proj_weight", + "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_138", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_243", + "inputs": [ + "linear_137", + "linear_138" + ], + "output": "mul_243", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_139", + "inputs": [ + "mul_243", + "p_model_model_language_model_layers_17_mlp_down_proj_weight", + "p_model_model_language_model_layers_17_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_139", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_191", + "inputs": [ + "add_188", + "linear_139" + ], + "output": "add_191", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_245", + "inputs": [ + "add_191", + "p_model_model_language_model_layers_18_input_layernorm_weight" + ], + "output": "mul_245", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_140", + "inputs": [ + "mul_245", + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_140", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_75", + "inputs": [ + "linear_140" + ], + "output": "transpose_75", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_141", + "inputs": [ + "mul_245", + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_141", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_142", + "inputs": [ + "mul_245", + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_142", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_143", + "inputs": [ + "mul_245", + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_143", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_31", + "inputs": [ + "states_36", + "transpose_75" + ], + "output": "cat_31", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_162", + "inputs": [ + "cat_31" + ], + "output": "slice_162", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_36" + }, + { + "node": "conv1d_14", + "inputs": [ + "cat_31", + "p_model_model_language_model_layers_18_linear_attn_conv1d_weight" + ], + "output": "conv1d_14", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_163", + "inputs": [ + "conv1d_14" + ], + "output": "slice_163", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_46", + "inputs": [ + "slice_163" + ], + "output": "silu_46", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_164", + "inputs": [ + "silu_46" + ], + "output": "slice_164", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_76", + "inputs": [ + "slice_164" + ], + "output": "transpose_76", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_165", + "inputs": [ + "transpose_76" + ], + "output": "slice_165", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_166", + "inputs": [ + "transpose_76" + ], + "output": "slice_166", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_167", + "inputs": [ + "transpose_76" + ], + "output": "slice_167", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_18", + "inputs": [ + "linear_142" + ], + "output": "sigmoid_18", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_14", + "inputs": [ + "p_model_model_language_model_layers_18_linear_attn_a_log" + ], + "output": "exp_14", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_22", + "inputs": [ + "exp_14" + ], + "output": "neg_22", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_194", + "inputs": [ + "linear_143", + "p_model_model_language_model_layers_18_linear_attn_dt_bias" + ], + "output": "add_194", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_14", + "inputs": [ + "add_194" + ], + "output": "softplus_14", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_246", + "inputs": [ + "neg_22", + "softplus_14" + ], + "output": "mul_246", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_247", + "inputs": [ + "slice_165" + ], + "output": "mul_247", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_29", + "inputs": [ + "mul_247" + ], + "output": "sum_29", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_195", + "inputs": [ + "sum_29" + ], + "output": "add_195", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_87", + "inputs": [ + "add_195" + ], + "output": "rsqrt_87", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_248", + "inputs": [ + "slice_165", + "rsqrt_87" + ], + "output": "mul_248", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_249", + "inputs": [ + "slice_166" + ], + "output": "mul_249", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_30", + "inputs": [ + "mul_249" + ], + "output": "sum_30", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_196", + "inputs": [ + "sum_30" + ], + "output": "add_196", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_88", + "inputs": [ + "add_196" + ], + "output": "rsqrt_88", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_250", + "inputs": [ + "slice_166", + "rsqrt_88" + ], + "output": "mul_250", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_14", + "inputs": [ + "mul_248", + "mul_250", + "slice_167", + "mul_246", + "sigmoid_18", + "states_37" + ], + "output": "normalized_gated_delta_rule_14", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_168", + "inputs": [ + "normalized_gated_delta_rule_14" + ], + "output": "slice_168", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_37" + }, + { + "node": "slice_169", + "inputs": [ + "normalized_gated_delta_rule_14" + ], + "output": "slice_169", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_77", + "inputs": [ + "slice_169" + ], + "output": "transpose_77", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_252", + "inputs": [ + "transpose_77", + "p_model_model_language_model_layers_18_linear_attn_norm_weight" + ], + "output": "mul_252", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_253", + "inputs": [ + "linear_141", + "mul_252" + ], + "output": "mul_253", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_144", + "inputs": [ + "mul_253", + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_144", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_198", + "inputs": [ + "add_191", + "linear_144" + ], + "output": "add_198", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_255", + "inputs": [ + "add_198", + "p_model_model_language_model_layers_18_post_attention_layernorm_weight" + ], + "output": "mul_255", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_145", + "inputs": [ + "mul_255", + "p_model_model_language_model_layers_18_mlp_gate_proj_weight", + "p_model_model_language_model_layers_18_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_145", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_146", + "inputs": [ + "mul_255", + "p_model_model_language_model_layers_18_mlp_up_proj_weight", + "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_146", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_256", + "inputs": [ + "linear_145", + "linear_146" + ], + "output": "mul_256", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_147", + "inputs": [ + "mul_256", + "p_model_model_language_model_layers_18_mlp_down_proj_weight", + "p_model_model_language_model_layers_18_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_147", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_201", + "inputs": [ + "add_198", + "linear_147" + ], + "output": "add_201", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_258", + "inputs": [ + "add_201", + "p_model_model_language_model_layers_19_input_layernorm_weight" + ], + "output": "mul_258", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_148", + "inputs": [ + "mul_258", + "p_model_model_language_model_layers_19_self_attn_q_proj_weight", + "p_model_model_language_model_layers_19_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_148", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/f7f0ee18ecdc712d36e487eef201cb442f28f824e5b6fac7e4d16cb6b7debec4.wgsl" + }, + { + "node": "slice_170", + "inputs": [ + "linear_148" + ], + "output": "slice_170", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2eeb54034621ff097879064a60fd4e0dd5d68201e48296ad234413a39579acc8.wgsl" + }, + { + "node": "slice_171", + "inputs": [ + "linear_148" + ], + "output": "slice_171", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0d1edb63bf4dc212d55f8a4c79a50e04c0d822bd9f6e193003e2071ba0d57d69.wgsl" + }, + { + "node": "mul_260", + "inputs": [ + "slice_170", + "p_model_model_language_model_layers_19_self_attn_q_norm_weight" + ], + "output": "mul_260", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl" + }, + { + "node": "transpose_78", + "inputs": [ + "mul_260" + ], + "output": "transpose_78", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/18a0b02a12ebf73203283d267342582d54ba411aeee9cbe21ad89ae890e4cfea.wgsl" + }, + { + "node": "linear_149", + "inputs": [ + "mul_258", + "p_model_model_language_model_layers_19_self_attn_k_proj_weight", + "p_model_model_language_model_layers_19_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_149", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl" + }, + { + "node": "mul_262", + "inputs": [ + "linear_149", + "p_model_model_language_model_layers_19_self_attn_k_norm_weight" + ], + "output": "mul_262", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/f994142054bcb47d6a554524dd8de9c3054467a1d07011027c47f475d0909c19.wgsl" + }, + { + "node": "transpose_79", + "inputs": [ + "mul_262" + ], + "output": "transpose_79", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl" + }, + { + "node": "linear_150", + "inputs": [ + "mul_258", + "p_model_model_language_model_layers_19_self_attn_v_proj_weight", + "p_model_model_language_model_layers_19_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_150", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl" + }, + { + "node": "transpose_80", + "inputs": [ + "linear_150" + ], + "output": "transpose_80", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl" + }, + { + "node": "slice_172", + "inputs": [ + "transpose_78" + ], + "output": "slice_172", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8b5e4c9b7889f2bccebf60987b6f5b8c23542c3475cd8d78656dc7ecffd81f8d.wgsl" + }, + { + "node": "slice_173", + "inputs": [ + "transpose_78" + ], + "output": "slice_173", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/f95724b4a3d2dbafaccd23a45f68234bb49e4a39168942060ca628960993f33d.wgsl" + }, + { + "node": "slice_174", + "inputs": [ + "transpose_79" + ], + "output": "slice_174", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/ad72f60d58133272228319f1c066e7cd43ac28573ddeeb7a6f69be01af341f57.wgsl" + }, + { + "node": "slice_175", + "inputs": [ + "transpose_79" + ], + "output": "slice_175", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/8af32f59dfb78c79ac2187ae707f0897b89fb8bb0e3d832e2edac308eefa603c.wgsl" + }, + { + "node": "mul_263", + "inputs": [ + "slice_172", + "mul" + ], + "output": "mul_263", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl" + }, + { + "node": "slice_176", + "inputs": [ + "slice_172" + ], + "output": "slice_176", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8ee286e06c1271162bea74e14b4bd5b1d66890648217c79435e6b3f1d4181ec8.wgsl" + }, + { + "node": "slice_177", + "inputs": [ + "slice_172" + ], + "output": "slice_177", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/80d22263b082805bc9d4499247f92b7a1a783a3e1cdd9c652d304492a1cfd425.wgsl" + }, + { + "node": "neg_23", + "inputs": [ + "slice_177" + ], + "output": "neg_23", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl" + }, + { + "node": "cat_32", + "inputs": [ + "neg_23", + "slice_176" + ], + "output": "cat_32", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/bda7ebd3bcf25c31938fca1a582e82b972ab742f85f826f085417b0e6f651e60.wgsl" + }, + { + "node": "mul_264", + "inputs": [ + "cat_32", + "mul_1" + ], + "output": "mul_264", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl" + }, + { + "node": "add_208", + "inputs": [ + "mul_263", + "mul_264" + ], + "output": "add_208", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl" + }, + { + "node": "mul_265", + "inputs": [ + "slice_174", + "mul" + ], + "output": "mul_265", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/d0be84d22b9e3b0cf73729e9a6fad1c6cf7d03fa01b45b8765ed70158578f2a9.wgsl" + }, + { + "node": "slice_178", + "inputs": [ + "slice_174" + ], + "output": "slice_178", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8c3a77238461b0ae13318bad504e1e0779f3f47d1782fe0ca78f20112c031146.wgsl" + }, + { + "node": "slice_179", + "inputs": [ + "slice_174" + ], + "output": "slice_179", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/b2ae1b161c8a6315f8863e47511607027dfe583aac1db4c1626371ddfa0c64e2.wgsl" + }, + { + "node": "neg_24", + "inputs": [ + "slice_179" + ], + "output": "neg_24", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/4fad29e3e337840e25132f10b3e3d55225f727ae13525ff0b0bdf31e03f09d74.wgsl" + }, + { + "node": "cat_33", + "inputs": [ + "neg_24", + "slice_178" + ], + "output": "cat_33", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/898086c99e03b2cdb78e3e5658c3fff202f1c87d09d8da08659bd1ba85e01b09.wgsl" + }, + { + "node": "mul_266", + "inputs": [ + "cat_33", + "mul_1" + ], + "output": "mul_266", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/d0be84d22b9e3b0cf73729e9a6fad1c6cf7d03fa01b45b8765ed70158578f2a9.wgsl" + }, + { + "node": "add_209", + "inputs": [ + "mul_265", + "mul_266" + ], + "output": "add_209", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/f5b51b1e0e52bd87a70a0ca169186ebb5734cbdc2b8f375ba0241485b1df4e66.wgsl" + }, + { + "node": "cat_34", + "inputs": [ + "add_208", + "slice_173" + ], + "output": "cat_34", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl" + }, + { + "node": "cat_35", + "inputs": [ + "add_209", + "slice_175" + ], + "output": "cat_35", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f276cf6656b999339df464b61934487f5c8eb7a0957cf6d902254ab1aa2ffcd0.wgsl" + }, + { + "node": "cache_update_8", + "inputs": [ + "cat_35", + "position_ids" + ], + "output": "cache_update_8", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_38", + "kernel": "kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl" + }, + { + "node": "cache_update_9", + "inputs": [ + "transpose_80", + "position_ids" + ], + "output": "cache_update_9", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_39", + "kernel": "kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl" + }, + { + "node": "matmul_10", + "inputs": [ + "cat_34", + "cache_update_8", + "cache_update_9", + "attention_mask" + ], + "output": "matmul_10", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f5b2ba528c9d245492ad94b337e8cafcc65deeb4a5e7975a55ca79cbfe0a29d4.wgsl" + }, + { + "node": "transpose_82", + "inputs": [ + "matmul_10" + ], + "output": "transpose_82", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2792ae5841828bd5af938c10096abe385e63bb3ec2a22e850f146739433c12bf.wgsl" + }, + { + "node": "sigmoid_19", + "inputs": [ + "slice_171" + ], + "output": "sigmoid_19", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a53c1ce51ebaec593ad667efae9603efe3807db8f0ee060a620ac031cc4f2b92.wgsl" + }, + { + "node": "mul_268", + "inputs": [ + "transpose_82", + "sigmoid_19" + ], + "output": "mul_268", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/45a1fd3ed514f208718651026bf2ee30dda66b02cdd950dc71e88eb6bdac0b10.wgsl" + }, + { + "node": "linear_151", + "inputs": [ + "mul_268", + "p_model_model_language_model_layers_19_self_attn_o_proj_weight", + "p_model_model_language_model_layers_19_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_151", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_211", + "inputs": [ + "add_201", + "linear_151" + ], + "output": "add_211", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_270", + "inputs": [ + "add_211", + "p_model_model_language_model_layers_19_post_attention_layernorm_weight" + ], + "output": "mul_270", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_152", + "inputs": [ + "mul_270", + "p_model_model_language_model_layers_19_mlp_gate_proj_weight", + "p_model_model_language_model_layers_19_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_152", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_153", + "inputs": [ + "mul_270", + "p_model_model_language_model_layers_19_mlp_up_proj_weight", + "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_153", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_271", + "inputs": [ + "linear_152", + "linear_153" + ], + "output": "mul_271", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_154", + "inputs": [ + "mul_271", + "p_model_model_language_model_layers_19_mlp_down_proj_weight", + "p_model_model_language_model_layers_19_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_154", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_214", + "inputs": [ + "add_211", + "linear_154" + ], + "output": "add_214", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_273", + "inputs": [ + "add_214", + "p_model_model_language_model_layers_20_input_layernorm_weight" + ], + "output": "mul_273", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_155", + "inputs": [ + "mul_273", + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_155", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_83", + "inputs": [ + "linear_155" + ], + "output": "transpose_83", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_156", + "inputs": [ + "mul_273", + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_156", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_157", + "inputs": [ + "mul_273", + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_157", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_158", + "inputs": [ + "mul_273", + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_158", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_36", + "inputs": [ + "states_40", + "transpose_83" + ], + "output": "cat_36", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_180", + "inputs": [ + "cat_36" + ], + "output": "slice_180", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_40" + }, + { + "node": "conv1d_15", + "inputs": [ + "cat_36", + "p_model_model_language_model_layers_20_linear_attn_conv1d_weight" + ], + "output": "conv1d_15", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_181", + "inputs": [ + "conv1d_15" + ], + "output": "slice_181", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_50", + "inputs": [ + "slice_181" + ], + "output": "silu_50", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_182", + "inputs": [ + "silu_50" + ], + "output": "slice_182", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_84", + "inputs": [ + "slice_182" + ], + "output": "transpose_84", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_183", + "inputs": [ + "transpose_84" + ], + "output": "slice_183", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_184", + "inputs": [ + "transpose_84" + ], + "output": "slice_184", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_185", + "inputs": [ + "transpose_84" + ], + "output": "slice_185", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_20", + "inputs": [ + "linear_157" + ], + "output": "sigmoid_20", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_15", + "inputs": [ + "p_model_model_language_model_layers_20_linear_attn_a_log" + ], + "output": "exp_15", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_25", + "inputs": [ + "exp_15" + ], + "output": "neg_25", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_217", + "inputs": [ + "linear_158", + "p_model_model_language_model_layers_20_linear_attn_dt_bias" + ], + "output": "add_217", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_15", + "inputs": [ + "add_217" + ], + "output": "softplus_15", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_274", + "inputs": [ + "neg_25", + "softplus_15" + ], + "output": "mul_274", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_275", + "inputs": [ + "slice_183" + ], + "output": "mul_275", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_31", + "inputs": [ + "mul_275" + ], + "output": "sum_31", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_218", + "inputs": [ + "sum_31" + ], + "output": "add_218", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_96", + "inputs": [ + "add_218" + ], + "output": "rsqrt_96", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_276", + "inputs": [ + "slice_183", + "rsqrt_96" + ], + "output": "mul_276", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_277", + "inputs": [ + "slice_184" + ], + "output": "mul_277", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_32", + "inputs": [ + "mul_277" + ], + "output": "sum_32", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_219", + "inputs": [ + "sum_32" + ], + "output": "add_219", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_97", + "inputs": [ + "add_219" + ], + "output": "rsqrt_97", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_278", + "inputs": [ + "slice_184", + "rsqrt_97" + ], + "output": "mul_278", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_15", + "inputs": [ + "mul_276", + "mul_278", + "slice_185", + "mul_274", + "sigmoid_20", + "states_41" + ], + "output": "normalized_gated_delta_rule_15", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_186", + "inputs": [ + "normalized_gated_delta_rule_15" + ], + "output": "slice_186", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_41" + }, + { + "node": "slice_187", + "inputs": [ + "normalized_gated_delta_rule_15" + ], + "output": "slice_187", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_85", + "inputs": [ + "slice_187" + ], + "output": "transpose_85", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_280", + "inputs": [ + "transpose_85", + "p_model_model_language_model_layers_20_linear_attn_norm_weight" + ], + "output": "mul_280", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_281", + "inputs": [ + "linear_156", + "mul_280" + ], + "output": "mul_281", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_159", + "inputs": [ + "mul_281", + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_159", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_221", + "inputs": [ + "add_214", + "linear_159" + ], + "output": "add_221", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_283", + "inputs": [ + "add_221", + "p_model_model_language_model_layers_20_post_attention_layernorm_weight" + ], + "output": "mul_283", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_160", + "inputs": [ + "mul_283", + "p_model_model_language_model_layers_20_mlp_gate_proj_weight", + "p_model_model_language_model_layers_20_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_160", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_161", + "inputs": [ + "mul_283", + "p_model_model_language_model_layers_20_mlp_up_proj_weight", + "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_161", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_284", + "inputs": [ + "linear_160", + "linear_161" + ], + "output": "mul_284", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_162", + "inputs": [ + "mul_284", + "p_model_model_language_model_layers_20_mlp_down_proj_weight", + "p_model_model_language_model_layers_20_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_162", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_224", + "inputs": [ + "add_221", + "linear_162" + ], + "output": "add_224", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_286", + "inputs": [ + "add_224", + "p_model_model_language_model_layers_21_input_layernorm_weight" + ], + "output": "mul_286", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_163", + "inputs": [ + "mul_286", + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_163", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_86", + "inputs": [ + "linear_163" + ], + "output": "transpose_86", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_164", + "inputs": [ + "mul_286", + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_164", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_165", + "inputs": [ + "mul_286", + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_165", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_166", + "inputs": [ + "mul_286", + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_166", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_37", + "inputs": [ + "states_42", + "transpose_86" + ], + "output": "cat_37", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_188", + "inputs": [ + "cat_37" + ], + "output": "slice_188", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_42" + }, + { + "node": "conv1d_16", + "inputs": [ + "cat_37", + "p_model_model_language_model_layers_21_linear_attn_conv1d_weight" + ], + "output": "conv1d_16", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_189", + "inputs": [ + "conv1d_16" + ], + "output": "slice_189", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_53", + "inputs": [ + "slice_189" + ], + "output": "silu_53", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_190", + "inputs": [ + "silu_53" + ], + "output": "slice_190", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_87", + "inputs": [ + "slice_190" + ], + "output": "transpose_87", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_191", + "inputs": [ + "transpose_87" + ], + "output": "slice_191", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_192", + "inputs": [ + "transpose_87" + ], + "output": "slice_192", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_193", + "inputs": [ + "transpose_87" + ], + "output": "slice_193", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_21", + "inputs": [ + "linear_165" + ], + "output": "sigmoid_21", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_16", + "inputs": [ + "p_model_model_language_model_layers_21_linear_attn_a_log" + ], + "output": "exp_16", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_26", + "inputs": [ + "exp_16" + ], + "output": "neg_26", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_227", + "inputs": [ + "linear_166", + "p_model_model_language_model_layers_21_linear_attn_dt_bias" + ], + "output": "add_227", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_16", + "inputs": [ + "add_227" + ], + "output": "softplus_16", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_287", + "inputs": [ + "neg_26", + "softplus_16" + ], + "output": "mul_287", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_288", + "inputs": [ + "slice_191" + ], + "output": "mul_288", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_33", + "inputs": [ + "mul_288" + ], + "output": "sum_33", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_228", + "inputs": [ + "sum_33" + ], + "output": "add_228", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_101", + "inputs": [ + "add_228" + ], + "output": "rsqrt_101", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_289", + "inputs": [ + "slice_191", + "rsqrt_101" + ], + "output": "mul_289", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_290", + "inputs": [ + "slice_192" + ], + "output": "mul_290", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_34", + "inputs": [ + "mul_290" + ], + "output": "sum_34", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_229", + "inputs": [ + "sum_34" + ], + "output": "add_229", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_102", + "inputs": [ + "add_229" + ], + "output": "rsqrt_102", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_291", + "inputs": [ + "slice_192", + "rsqrt_102" + ], + "output": "mul_291", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_16", + "inputs": [ + "mul_289", + "mul_291", + "slice_193", + "mul_287", + "sigmoid_21", + "states_43" + ], + "output": "normalized_gated_delta_rule_16", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_194", + "inputs": [ + "normalized_gated_delta_rule_16" + ], + "output": "slice_194", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_43" + }, + { + "node": "slice_195", + "inputs": [ + "normalized_gated_delta_rule_16" + ], + "output": "slice_195", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_88", + "inputs": [ + "slice_195" + ], + "output": "transpose_88", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_293", + "inputs": [ + "transpose_88", + "p_model_model_language_model_layers_21_linear_attn_norm_weight" + ], + "output": "mul_293", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_294", + "inputs": [ + "linear_164", + "mul_293" + ], + "output": "mul_294", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_167", + "inputs": [ + "mul_294", + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_167", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_231", + "inputs": [ + "add_224", + "linear_167" + ], + "output": "add_231", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_296", + "inputs": [ + "add_231", + "p_model_model_language_model_layers_21_post_attention_layernorm_weight" + ], + "output": "mul_296", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_168", + "inputs": [ + "mul_296", + "p_model_model_language_model_layers_21_mlp_gate_proj_weight", + "p_model_model_language_model_layers_21_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_168", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_169", + "inputs": [ + "mul_296", + "p_model_model_language_model_layers_21_mlp_up_proj_weight", + "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_169", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_297", + "inputs": [ + "linear_168", + "linear_169" + ], + "output": "mul_297", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_170", + "inputs": [ + "mul_297", + "p_model_model_language_model_layers_21_mlp_down_proj_weight", + "p_model_model_language_model_layers_21_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_170", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_234", + "inputs": [ + "add_231", + "linear_170" + ], + "output": "add_234", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_299", + "inputs": [ + "add_234", + "p_model_model_language_model_layers_22_input_layernorm_weight" + ], + "output": "mul_299", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_171", + "inputs": [ + "mul_299", + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_171", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl" + }, + { + "node": "transpose_89", + "inputs": [ + "linear_171" + ], + "output": "transpose_89", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl" + }, + { + "node": "linear_172", + "inputs": [ + "mul_299", + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_172", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl" + }, + { + "node": "linear_173", + "inputs": [ + "mul_299", + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_173", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "linear_174", + "inputs": [ + "mul_299", + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_174", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl" + }, + { + "node": "cat_38", + "inputs": [ + "states_44", + "transpose_89" + ], + "output": "cat_38", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl" + }, + { + "node": "slice_196", + "inputs": [ + "cat_38" + ], + "output": "slice_196", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl", + "inPlaceOf": "states_44" + }, + { + "node": "conv1d_17", + "inputs": [ + "cat_38", + "p_model_model_language_model_layers_22_linear_attn_conv1d_weight" + ], + "output": "conv1d_17", + "workgroups": [ + 1056, + 1, + 1 + ], + "kernel": "kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl" + }, + { + "node": "slice_197", + "inputs": [ + "conv1d_17" + ], + "output": "slice_197", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl" + }, + { + "node": "silu_56", + "inputs": [ + "slice_197" + ], + "output": "silu_56", + "workgroups": [ + 768, + 1, + 1 + ], + "kernel": "kernels/b6c3ce0defd66eef66dc3fd6eb9eab3fee12d179332698c3c764009988628b49.wgsl" + }, + { + "node": "slice_198", + "inputs": [ + "silu_56" + ], + "output": "slice_198", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl" + }, + { + "node": "transpose_90", + "inputs": [ + "slice_198" + ], + "output": "transpose_90", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/fa90a055381629df20c59052b5e44670510dd1666b9f2b06a554dfed9a002953.wgsl" + }, + { + "node": "slice_199", + "inputs": [ + "transpose_90" + ], + "output": "slice_199", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e74eaedc2b731f96007caaea3db1518767845c24bc917b9c2b71553e22139130.wgsl" + }, + { + "node": "slice_200", + "inputs": [ + "transpose_90" + ], + "output": "slice_200", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a9f0eef095f015e2099adf81552a6727c45e70fc11f5e50b777d36b6370ab986.wgsl" + }, + { + "node": "slice_201", + "inputs": [ + "transpose_90" + ], + "output": "slice_201", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl" + }, + { + "node": "sigmoid_22", + "inputs": [ + "linear_173" + ], + "output": "sigmoid_22", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f233dea95d020538ac476af9846f67cba0d137fd562d5f652cb333f0b3273c76.wgsl" + }, + { + "node": "exp_17", + "inputs": [ + "p_model_model_language_model_layers_22_linear_attn_a_log" + ], + "output": "exp_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_27", + "inputs": [ + "exp_17" + ], + "output": "neg_27", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_237", + "inputs": [ + "linear_174", + "p_model_model_language_model_layers_22_linear_attn_dt_bias" + ], + "output": "add_237", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl" + }, + { + "node": "softplus_17", + "inputs": [ + "add_237" + ], + "output": "softplus_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl" + }, + { + "node": "mul_300", + "inputs": [ + "neg_27", + "softplus_17" + ], + "output": "mul_300", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/f66b0b2ea4229895bb6599fb5d3528be17eb9020c19829b54dd0fbb4d4857039.wgsl" + }, + { + "node": "mul_301", + "inputs": [ + "slice_199" + ], + "output": "mul_301", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_35", + "inputs": [ + "mul_301" + ], + "output": "sum_35", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_238", + "inputs": [ + "sum_35" + ], + "output": "add_238", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_106", + "inputs": [ + "add_238" + ], + "output": "rsqrt_106", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_302", + "inputs": [ + "slice_199", + "rsqrt_106" + ], + "output": "mul_302", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "mul_303", + "inputs": [ + "slice_200" + ], + "output": "mul_303", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/f45156a1c571e56cf0730c88fcd58dbb5716b2cea086ebb9e0dfedfe5dcbf658.wgsl" + }, + { + "node": "sum_36", + "inputs": [ + "mul_303" + ], + "output": "sum_36", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/c1096000361a304457865401d6266a640469ca04c0bbfe2eaa657e656b2fca33.wgsl" + }, + { + "node": "add_239", + "inputs": [ + "sum_36" + ], + "output": "add_239", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl" + }, + { + "node": "rsqrt_107", + "inputs": [ + "add_239" + ], + "output": "rsqrt_107", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/de4b2b6e4d6c225d58c25261df5c8f73711c3bbbd04a31d1ea5c5bffc827c6e9.wgsl" + }, + { + "node": "mul_304", + "inputs": [ + "slice_200", + "rsqrt_107" + ], + "output": "mul_304", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/c3735f34cafbac42efa5928641cefdaf7277659cc2675a36756bba97240ba5ae.wgsl" + }, + { + "node": "normalized_gated_delta_rule_17", + "inputs": [ + "mul_302", + "mul_304", + "slice_201", + "mul_300", + "sigmoid_22", + "states_45" + ], + "output": "normalized_gated_delta_rule_17", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl" + }, + { + "node": "slice_202", + "inputs": [ + "normalized_gated_delta_rule_17" + ], + "output": "slice_202", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl", + "inPlaceOf": "states_45" + }, + { + "node": "slice_203", + "inputs": [ + "normalized_gated_delta_rule_17" + ], + "output": "slice_203", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/e91fae5817b3b5e4b0025d4fa70d4cfa10b010d038a3f45c6348d25379df287d.wgsl" + }, + { + "node": "transpose_91", + "inputs": [ + "slice_203" + ], + "output": "transpose_91", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/fa7c163b443e5f494b83f24da518fcad3e353d6e85facc3e3478f6005f964bd6.wgsl" + }, + { + "node": "mul_306", + "inputs": [ + "transpose_91", + "p_model_model_language_model_layers_22_linear_attn_norm_weight" + ], + "output": "mul_306", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl" + }, + { + "node": "mul_307", + "inputs": [ + "linear_172", + "mul_306" + ], + "output": "mul_307", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl" + }, + { + "node": "linear_175", + "inputs": [ + "mul_307", + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_175", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_241", + "inputs": [ + "add_234", + "linear_175" + ], + "output": "add_241", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_309", + "inputs": [ + "add_241", + "p_model_model_language_model_layers_22_post_attention_layernorm_weight" + ], + "output": "mul_309", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_176", + "inputs": [ + "mul_309", + "p_model_model_language_model_layers_22_mlp_gate_proj_weight", + "p_model_model_language_model_layers_22_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_176", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_177", + "inputs": [ + "mul_309", + "p_model_model_language_model_layers_22_mlp_up_proj_weight", + "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_177", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_310", + "inputs": [ + "linear_176", + "linear_177" + ], + "output": "mul_310", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_178", + "inputs": [ + "mul_310", + "p_model_model_language_model_layers_22_mlp_down_proj_weight", + "p_model_model_language_model_layers_22_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_178", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_244", + "inputs": [ + "add_241", + "linear_178" + ], + "output": "add_244", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_312", + "inputs": [ + "add_244", + "p_model_model_language_model_layers_23_input_layernorm_weight" + ], + "output": "mul_312", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_179", + "inputs": [ + "mul_312", + "p_model_model_language_model_layers_23_self_attn_q_proj_weight", + "p_model_model_language_model_layers_23_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_179", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/f7f0ee18ecdc712d36e487eef201cb442f28f824e5b6fac7e4d16cb6b7debec4.wgsl" + }, + { + "node": "slice_204", + "inputs": [ + "linear_179" + ], + "output": "slice_204", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2eeb54034621ff097879064a60fd4e0dd5d68201e48296ad234413a39579acc8.wgsl" + }, + { + "node": "slice_205", + "inputs": [ + "linear_179" + ], + "output": "slice_205", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0d1edb63bf4dc212d55f8a4c79a50e04c0d822bd9f6e193003e2071ba0d57d69.wgsl" + }, + { + "node": "mul_314", + "inputs": [ + "slice_204", + "p_model_model_language_model_layers_23_self_attn_q_norm_weight" + ], + "output": "mul_314", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl" + }, + { + "node": "transpose_92", + "inputs": [ + "mul_314" + ], + "output": "transpose_92", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/18a0b02a12ebf73203283d267342582d54ba411aeee9cbe21ad89ae890e4cfea.wgsl" + }, + { + "node": "linear_180", + "inputs": [ + "mul_312", + "p_model_model_language_model_layers_23_self_attn_k_proj_weight", + "p_model_model_language_model_layers_23_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_180", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl" + }, + { + "node": "mul_316", + "inputs": [ + "linear_180", + "p_model_model_language_model_layers_23_self_attn_k_norm_weight" + ], + "output": "mul_316", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/f994142054bcb47d6a554524dd8de9c3054467a1d07011027c47f475d0909c19.wgsl" + }, + { + "node": "transpose_93", + "inputs": [ + "mul_316" + ], + "output": "transpose_93", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl" + }, + { + "node": "linear_181", + "inputs": [ + "mul_312", + "p_model_model_language_model_layers_23_self_attn_v_proj_weight", + "p_model_model_language_model_layers_23_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_181", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl" + }, + { + "node": "transpose_94", + "inputs": [ + "linear_181" + ], + "output": "transpose_94", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl" + }, + { + "node": "slice_206", + "inputs": [ + "transpose_92" + ], + "output": "slice_206", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/8b5e4c9b7889f2bccebf60987b6f5b8c23542c3475cd8d78656dc7ecffd81f8d.wgsl" + }, + { + "node": "slice_207", + "inputs": [ + "transpose_92" + ], + "output": "slice_207", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/f95724b4a3d2dbafaccd23a45f68234bb49e4a39168942060ca628960993f33d.wgsl" + }, + { + "node": "slice_208", + "inputs": [ + "transpose_93" + ], + "output": "slice_208", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/ad72f60d58133272228319f1c066e7cd43ac28573ddeeb7a6f69be01af341f57.wgsl" + }, + { + "node": "slice_209", + "inputs": [ + "transpose_93" + ], + "output": "slice_209", + "workgroups": [ + 24, + 1, + 1 + ], + "kernel": "kernels/8af32f59dfb78c79ac2187ae707f0897b89fb8bb0e3d832e2edac308eefa603c.wgsl" + }, + { + "node": "mul_317", + "inputs": [ + "slice_206", + "mul" + ], + "output": "mul_317", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl" + }, + { + "node": "slice_210", + "inputs": [ + "slice_206" + ], + "output": "slice_210", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/8ee286e06c1271162bea74e14b4bd5b1d66890648217c79435e6b3f1d4181ec8.wgsl" + }, + { + "node": "slice_211", + "inputs": [ + "slice_206" + ], + "output": "slice_211", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/80d22263b082805bc9d4499247f92b7a1a783a3e1cdd9c652d304492a1cfd425.wgsl" + }, + { + "node": "neg_28", + "inputs": [ + "slice_211" + ], + "output": "neg_28", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl" + }, + { + "node": "cat_39", + "inputs": [ + "neg_28", + "slice_210" + ], + "output": "cat_39", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/bda7ebd3bcf25c31938fca1a582e82b972ab742f85f826f085417b0e6f651e60.wgsl" + }, + { + "node": "mul_318", + "inputs": [ + "cat_39", + "mul_1" + ], + "output": "mul_318", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl" + }, + { + "node": "add_251", + "inputs": [ + "mul_317", + "mul_318" + ], + "output": "add_251", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl" + }, + { + "node": "mul_319", + "inputs": [ + "slice_208", + "mul" + ], + "output": "mul_319", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/d0be84d22b9e3b0cf73729e9a6fad1c6cf7d03fa01b45b8765ed70158578f2a9.wgsl" + }, + { + "node": "slice_212", + "inputs": [ + "slice_208" + ], + "output": "slice_212", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/8c3a77238461b0ae13318bad504e1e0779f3f47d1782fe0ca78f20112c031146.wgsl" + }, + { + "node": "slice_213", + "inputs": [ + "slice_208" + ], + "output": "slice_213", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/b2ae1b161c8a6315f8863e47511607027dfe583aac1db4c1626371ddfa0c64e2.wgsl" + }, + { + "node": "neg_29", + "inputs": [ + "slice_213" + ], + "output": "neg_29", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/4fad29e3e337840e25132f10b3e3d55225f727ae13525ff0b0bdf31e03f09d74.wgsl" + }, + { + "node": "cat_40", + "inputs": [ + "neg_29", + "slice_212" + ], + "output": "cat_40", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/898086c99e03b2cdb78e3e5658c3fff202f1c87d09d8da08659bd1ba85e01b09.wgsl" + }, + { + "node": "mul_320", + "inputs": [ + "cat_40", + "mul_1" + ], + "output": "mul_320", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/d0be84d22b9e3b0cf73729e9a6fad1c6cf7d03fa01b45b8765ed70158578f2a9.wgsl" + }, + { + "node": "add_252", + "inputs": [ + "mul_319", + "mul_320" + ], + "output": "add_252", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/f5b51b1e0e52bd87a70a0ca169186ebb5734cbdc2b8f375ba0241485b1df4e66.wgsl" + }, + { + "node": "cat_41", + "inputs": [ + "add_251", + "slice_207" + ], + "output": "cat_41", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl" + }, + { + "node": "cat_42", + "inputs": [ + "add_252", + "slice_209" + ], + "output": "cat_42", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f276cf6656b999339df464b61934487f5c8eb7a0957cf6d902254ab1aa2ffcd0.wgsl" + }, + { + "node": "cache_update_10", + "inputs": [ + "cat_42", + "position_ids" + ], + "output": "cache_update_10", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_46", + "kernel": "kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl" + }, + { + "node": "cache_update_11", + "inputs": [ + "transpose_94", + "position_ids" + ], + "output": "cache_update_11", + "workgroups": [ + 32, + 1, + 1 + ], + "inPlaceOf": "states_47", + "kernel": "kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl" + }, + { + "node": "matmul_12", + "inputs": [ + "cat_41", + "cache_update_10", + "cache_update_11", + "attention_mask" + ], + "output": "matmul_12", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f5b2ba528c9d245492ad94b337e8cafcc65deeb4a5e7975a55ca79cbfe0a29d4.wgsl" + }, + { + "node": "transpose_96", + "inputs": [ + "matmul_12" + ], + "output": "transpose_96", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/2792ae5841828bd5af938c10096abe385e63bb3ec2a22e850f146739433c12bf.wgsl" + }, + { + "node": "sigmoid_23", + "inputs": [ + "slice_205" + ], + "output": "sigmoid_23", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/a53c1ce51ebaec593ad667efae9603efe3807db8f0ee060a620ac031cc4f2b92.wgsl" + }, + { + "node": "mul_322", + "inputs": [ + "transpose_96", + "sigmoid_23" + ], + "output": "mul_322", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/45a1fd3ed514f208718651026bf2ee30dda66b02cdd950dc71e88eb6bdac0b10.wgsl" + }, + { + "node": "linear_182", + "inputs": [ + "mul_322", + "p_model_model_language_model_layers_23_self_attn_o_proj_weight", + "p_model_model_language_model_layers_23_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_182", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/bb3e27e020bc22a0c8c21a5f9910a0a26b73ace5f68d9d49d1e4d67d53232060.wgsl" + }, + { + "node": "add_254", + "inputs": [ + "add_244", + "linear_182" + ], + "output": "add_254", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_324", + "inputs": [ + "add_254", + "p_model_model_language_model_layers_23_post_attention_layernorm_weight" + ], + "output": "mul_324", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_183", + "inputs": [ + "mul_324", + "p_model_model_language_model_layers_23_mlp_gate_proj_weight", + "p_model_model_language_model_layers_23_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_183", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "linear_184", + "inputs": [ + "mul_324", + "p_model_model_language_model_layers_23_mlp_up_proj_weight", + "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_184", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/f4da95f36c7aa6394d358a8d33a33ebeb3e404dbe56ddb554bad895ac538e5ad.wgsl" + }, + { + "node": "mul_325", + "inputs": [ + "linear_183", + "linear_184" + ], + "output": "mul_325", + "workgroups": [ + 224, + 1, + 1 + ], + "kernel": "kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl" + }, + { + "node": "linear_185", + "inputs": [ + "mul_325", + "p_model_model_language_model_layers_23_mlp_down_proj_weight", + "p_model_model_language_model_layers_23_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_185", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl" + }, + { + "node": "add_257", + "inputs": [ + "add_254", + "linear_185" + ], + "output": "add_257", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/c93686edb422c5385c7a6f484e76be0aa5cfa478cc80bc900a707e19fa2de99c.wgsl" + }, + { + "node": "mul_327", + "inputs": [ + "add_257", + "p_model_model_language_model_norm_weight" + ], + "output": "mul_327", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl" + }, + { + "node": "linear_186__part_0", + "inputs": [ + "mul_327", + "p_model_model_language_model_embed_tokens_weight__rows_0_131072", + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072" + ], + "output": "linear_186__part_0", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/adfbba60ade7143599d7de0faa4904afc1400ccc3cf84602f22a2fb004d0398e.wgsl" + }, + { + "node": "linear_186__part_1", + "inputs": [ + "mul_327", + "p_model_model_language_model_embed_tokens_weight__rows_131072_117248", + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248" + ], + "output": "linear_186__part_1", + "workgroups": [ + 1832, + 1, + 1 + ], + "kernel": "kernels/0153b6b641515b31ea2980613978c809e225435001b48fbec69afa520251a7af.wgsl" + }, + { + "node": "linear_186", + "inputs": [ + "linear_186__part_0", + "linear_186__part_1" + ], + "output": "linear_186", + "workgroups": [ + 15520, + 1, + 1 + ], + "kernel": "kernels/781d08cc08a9e49246e543d15b3d95775985fd130f5536c36bac80b33d2b86da.wgsl" + } + ], + "memory": { + "slots": [ + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 128, + 16, + 16, + 65536, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 134217728, + 16777216, + 120061952, + 15007744, + 16384, + 16384, + 16384, + 48, + 512, + 512, + 512, + 512, + 1024, + 98304, + 98304, + 196608, + 270336, + 32768, + 1081344, + 4096, + 4096, + 8192, + 2097152, + 1875968, + 3973120 + ], + "assignment": { + "p_model_model_language_model_layers_0_input_layernorm_weight": 0, + "p_model_model_language_model_layers_0_linear_attn_a_log": 1, + "p_model_model_language_model_layers_0_linear_attn_dt_bias": 2, + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight": 3, + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight": 4, + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight": 5, + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight": 6, + "p_model_model_language_model_layers_0_linear_attn_conv1d_weight": 7, + "p_model_model_language_model_layers_0_linear_attn_norm_weight": 8, + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight": 9, + "p_model_model_language_model_layers_0_post_attention_layernorm_weight": 10, + "p_model_model_language_model_layers_0_mlp_gate_proj_weight": 11, + "p_model_model_language_model_layers_0_mlp_up_proj_weight": 12, + "p_model_model_language_model_layers_0_mlp_down_proj_weight": 13, + "p_model_model_language_model_layers_1_input_layernorm_weight": 14, + "p_model_model_language_model_layers_1_linear_attn_a_log": 15, + "p_model_model_language_model_layers_1_linear_attn_dt_bias": 16, + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight": 17, + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight": 18, + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight": 19, + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight": 20, + "p_model_model_language_model_layers_1_linear_attn_conv1d_weight": 21, + "p_model_model_language_model_layers_1_linear_attn_norm_weight": 22, + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight": 23, + "p_model_model_language_model_layers_1_post_attention_layernorm_weight": 24, + "p_model_model_language_model_layers_1_mlp_gate_proj_weight": 25, + "p_model_model_language_model_layers_1_mlp_up_proj_weight": 26, + "p_model_model_language_model_layers_1_mlp_down_proj_weight": 27, + "p_model_model_language_model_layers_2_input_layernorm_weight": 28, + "p_model_model_language_model_layers_2_linear_attn_a_log": 29, + "p_model_model_language_model_layers_2_linear_attn_dt_bias": 30, + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight": 31, + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight": 32, + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight": 33, + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight": 34, + "p_model_model_language_model_layers_2_linear_attn_conv1d_weight": 35, + "p_model_model_language_model_layers_2_linear_attn_norm_weight": 36, + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight": 37, + "p_model_model_language_model_layers_2_post_attention_layernorm_weight": 38, + "p_model_model_language_model_layers_2_mlp_gate_proj_weight": 39, + "p_model_model_language_model_layers_2_mlp_up_proj_weight": 40, + "p_model_model_language_model_layers_2_mlp_down_proj_weight": 41, + "p_model_model_language_model_layers_3_input_layernorm_weight": 42, + "p_model_model_language_model_layers_3_self_attn_q_proj_weight": 43, + "p_model_model_language_model_layers_3_self_attn_q_norm_weight": 44, + "p_model_model_language_model_layers_3_self_attn_k_proj_weight": 45, + "p_model_model_language_model_layers_3_self_attn_k_norm_weight": 46, + "p_model_model_language_model_layers_3_self_attn_v_proj_weight": 47, + "p_model_model_language_model_layers_3_self_attn_o_proj_weight": 48, + "p_model_model_language_model_layers_3_post_attention_layernorm_weight": 49, + "p_model_model_language_model_layers_3_mlp_gate_proj_weight": 50, + "p_model_model_language_model_layers_3_mlp_up_proj_weight": 51, + "p_model_model_language_model_layers_3_mlp_down_proj_weight": 52, + "p_model_model_language_model_layers_4_input_layernorm_weight": 53, + "p_model_model_language_model_layers_4_linear_attn_a_log": 54, + "p_model_model_language_model_layers_4_linear_attn_dt_bias": 55, + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight": 56, + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight": 57, + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight": 58, + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight": 59, + "p_model_model_language_model_layers_4_linear_attn_conv1d_weight": 60, + "p_model_model_language_model_layers_4_linear_attn_norm_weight": 61, + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight": 62, + "p_model_model_language_model_layers_4_post_attention_layernorm_weight": 63, + "p_model_model_language_model_layers_4_mlp_gate_proj_weight": 64, + "p_model_model_language_model_layers_4_mlp_up_proj_weight": 65, + "p_model_model_language_model_layers_4_mlp_down_proj_weight": 66, + "p_model_model_language_model_layers_5_input_layernorm_weight": 67, + "p_model_model_language_model_layers_5_linear_attn_a_log": 68, + "p_model_model_language_model_layers_5_linear_attn_dt_bias": 69, + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight": 70, + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight": 71, + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight": 72, + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight": 73, + "p_model_model_language_model_layers_5_linear_attn_conv1d_weight": 74, + "p_model_model_language_model_layers_5_linear_attn_norm_weight": 75, + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight": 76, + "p_model_model_language_model_layers_5_post_attention_layernorm_weight": 77, + "p_model_model_language_model_layers_5_mlp_gate_proj_weight": 78, + "p_model_model_language_model_layers_5_mlp_up_proj_weight": 79, + "p_model_model_language_model_layers_5_mlp_down_proj_weight": 80, + "p_model_model_language_model_layers_6_input_layernorm_weight": 81, + "p_model_model_language_model_layers_6_linear_attn_a_log": 82, + "p_model_model_language_model_layers_6_linear_attn_dt_bias": 83, + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight": 84, + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight": 85, + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight": 86, + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight": 87, + "p_model_model_language_model_layers_6_linear_attn_conv1d_weight": 88, + "p_model_model_language_model_layers_6_linear_attn_norm_weight": 89, + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight": 90, + "p_model_model_language_model_layers_6_post_attention_layernorm_weight": 91, + "p_model_model_language_model_layers_6_mlp_gate_proj_weight": 92, + "p_model_model_language_model_layers_6_mlp_up_proj_weight": 93, + "p_model_model_language_model_layers_6_mlp_down_proj_weight": 94, + "p_model_model_language_model_layers_7_input_layernorm_weight": 95, + "p_model_model_language_model_layers_7_self_attn_q_proj_weight": 96, + "p_model_model_language_model_layers_7_self_attn_q_norm_weight": 97, + "p_model_model_language_model_layers_7_self_attn_k_proj_weight": 98, + "p_model_model_language_model_layers_7_self_attn_k_norm_weight": 99, + "p_model_model_language_model_layers_7_self_attn_v_proj_weight": 100, + "p_model_model_language_model_layers_7_self_attn_o_proj_weight": 101, + "p_model_model_language_model_layers_7_post_attention_layernorm_weight": 102, + "p_model_model_language_model_layers_7_mlp_gate_proj_weight": 103, + "p_model_model_language_model_layers_7_mlp_up_proj_weight": 104, + "p_model_model_language_model_layers_7_mlp_down_proj_weight": 105, + "p_model_model_language_model_layers_8_input_layernorm_weight": 106, + "p_model_model_language_model_layers_8_linear_attn_a_log": 107, + "p_model_model_language_model_layers_8_linear_attn_dt_bias": 108, + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight": 109, + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight": 110, + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight": 111, + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight": 112, + "p_model_model_language_model_layers_8_linear_attn_conv1d_weight": 113, + "p_model_model_language_model_layers_8_linear_attn_norm_weight": 114, + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight": 115, + "p_model_model_language_model_layers_8_post_attention_layernorm_weight": 116, + "p_model_model_language_model_layers_8_mlp_gate_proj_weight": 117, + "p_model_model_language_model_layers_8_mlp_up_proj_weight": 118, + "p_model_model_language_model_layers_8_mlp_down_proj_weight": 119, + "p_model_model_language_model_layers_9_input_layernorm_weight": 120, + "p_model_model_language_model_layers_9_linear_attn_a_log": 121, + "p_model_model_language_model_layers_9_linear_attn_dt_bias": 122, + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight": 123, + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight": 124, + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight": 125, + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight": 126, + "p_model_model_language_model_layers_9_linear_attn_conv1d_weight": 127, + "p_model_model_language_model_layers_9_linear_attn_norm_weight": 128, + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight": 129, + "p_model_model_language_model_layers_9_post_attention_layernorm_weight": 130, + "p_model_model_language_model_layers_9_mlp_gate_proj_weight": 131, + "p_model_model_language_model_layers_9_mlp_up_proj_weight": 132, + "p_model_model_language_model_layers_9_mlp_down_proj_weight": 133, + "p_model_model_language_model_layers_10_input_layernorm_weight": 134, + "p_model_model_language_model_layers_10_linear_attn_a_log": 135, + "p_model_model_language_model_layers_10_linear_attn_dt_bias": 136, + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight": 137, + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight": 138, + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight": 139, + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight": 140, + "p_model_model_language_model_layers_10_linear_attn_conv1d_weight": 141, + "p_model_model_language_model_layers_10_linear_attn_norm_weight": 142, + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight": 143, + "p_model_model_language_model_layers_10_post_attention_layernorm_weight": 144, + "p_model_model_language_model_layers_10_mlp_gate_proj_weight": 145, + "p_model_model_language_model_layers_10_mlp_up_proj_weight": 146, + "p_model_model_language_model_layers_10_mlp_down_proj_weight": 147, + "p_model_model_language_model_layers_11_input_layernorm_weight": 148, + "p_model_model_language_model_layers_11_self_attn_q_proj_weight": 149, + "p_model_model_language_model_layers_11_self_attn_q_norm_weight": 150, + "p_model_model_language_model_layers_11_self_attn_k_proj_weight": 151, + "p_model_model_language_model_layers_11_self_attn_k_norm_weight": 152, + "p_model_model_language_model_layers_11_self_attn_v_proj_weight": 153, + "p_model_model_language_model_layers_11_self_attn_o_proj_weight": 154, + "p_model_model_language_model_layers_11_post_attention_layernorm_weight": 155, + "p_model_model_language_model_layers_11_mlp_gate_proj_weight": 156, + "p_model_model_language_model_layers_11_mlp_up_proj_weight": 157, + "p_model_model_language_model_layers_11_mlp_down_proj_weight": 158, + "p_model_model_language_model_layers_12_input_layernorm_weight": 159, + "p_model_model_language_model_layers_12_linear_attn_a_log": 160, + "p_model_model_language_model_layers_12_linear_attn_dt_bias": 161, + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight": 162, + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight": 163, + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight": 164, + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight": 165, + "p_model_model_language_model_layers_12_linear_attn_conv1d_weight": 166, + "p_model_model_language_model_layers_12_linear_attn_norm_weight": 167, + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight": 168, + "p_model_model_language_model_layers_12_post_attention_layernorm_weight": 169, + "p_model_model_language_model_layers_12_mlp_gate_proj_weight": 170, + "p_model_model_language_model_layers_12_mlp_up_proj_weight": 171, + "p_model_model_language_model_layers_12_mlp_down_proj_weight": 172, + "p_model_model_language_model_layers_13_input_layernorm_weight": 173, + "p_model_model_language_model_layers_13_linear_attn_a_log": 174, + "p_model_model_language_model_layers_13_linear_attn_dt_bias": 175, + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight": 176, + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight": 177, + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight": 178, + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight": 179, + "p_model_model_language_model_layers_13_linear_attn_conv1d_weight": 180, + "p_model_model_language_model_layers_13_linear_attn_norm_weight": 181, + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight": 182, + "p_model_model_language_model_layers_13_post_attention_layernorm_weight": 183, + "p_model_model_language_model_layers_13_mlp_gate_proj_weight": 184, + "p_model_model_language_model_layers_13_mlp_up_proj_weight": 185, + "p_model_model_language_model_layers_13_mlp_down_proj_weight": 186, + "p_model_model_language_model_layers_14_input_layernorm_weight": 187, + "p_model_model_language_model_layers_14_linear_attn_a_log": 188, + "p_model_model_language_model_layers_14_linear_attn_dt_bias": 189, + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight": 190, + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight": 191, + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight": 192, + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight": 193, + "p_model_model_language_model_layers_14_linear_attn_conv1d_weight": 194, + "p_model_model_language_model_layers_14_linear_attn_norm_weight": 195, + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight": 196, + "p_model_model_language_model_layers_14_post_attention_layernorm_weight": 197, + "p_model_model_language_model_layers_14_mlp_gate_proj_weight": 198, + "p_model_model_language_model_layers_14_mlp_up_proj_weight": 199, + "p_model_model_language_model_layers_14_mlp_down_proj_weight": 200, + "p_model_model_language_model_layers_15_input_layernorm_weight": 201, + "p_model_model_language_model_layers_15_self_attn_q_proj_weight": 202, + "p_model_model_language_model_layers_15_self_attn_q_norm_weight": 203, + "p_model_model_language_model_layers_15_self_attn_k_proj_weight": 204, + "p_model_model_language_model_layers_15_self_attn_k_norm_weight": 205, + "p_model_model_language_model_layers_15_self_attn_v_proj_weight": 206, + "p_model_model_language_model_layers_15_self_attn_o_proj_weight": 207, + "p_model_model_language_model_layers_15_post_attention_layernorm_weight": 208, + "p_model_model_language_model_layers_15_mlp_gate_proj_weight": 209, + "p_model_model_language_model_layers_15_mlp_up_proj_weight": 210, + "p_model_model_language_model_layers_15_mlp_down_proj_weight": 211, + "p_model_model_language_model_layers_16_input_layernorm_weight": 212, + "p_model_model_language_model_layers_16_linear_attn_a_log": 213, + "p_model_model_language_model_layers_16_linear_attn_dt_bias": 214, + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight": 215, + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight": 216, + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight": 217, + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight": 218, + "p_model_model_language_model_layers_16_linear_attn_conv1d_weight": 219, + "p_model_model_language_model_layers_16_linear_attn_norm_weight": 220, + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight": 221, + "p_model_model_language_model_layers_16_post_attention_layernorm_weight": 222, + "p_model_model_language_model_layers_16_mlp_gate_proj_weight": 223, + "p_model_model_language_model_layers_16_mlp_up_proj_weight": 224, + "p_model_model_language_model_layers_16_mlp_down_proj_weight": 225, + "p_model_model_language_model_layers_17_input_layernorm_weight": 226, + "p_model_model_language_model_layers_17_linear_attn_a_log": 227, + "p_model_model_language_model_layers_17_linear_attn_dt_bias": 228, + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight": 229, + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight": 230, + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight": 231, + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight": 232, + "p_model_model_language_model_layers_17_linear_attn_conv1d_weight": 233, + "p_model_model_language_model_layers_17_linear_attn_norm_weight": 234, + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight": 235, + "p_model_model_language_model_layers_17_post_attention_layernorm_weight": 236, + "p_model_model_language_model_layers_17_mlp_gate_proj_weight": 237, + "p_model_model_language_model_layers_17_mlp_up_proj_weight": 238, + "p_model_model_language_model_layers_17_mlp_down_proj_weight": 239, + "p_model_model_language_model_layers_18_input_layernorm_weight": 240, + "p_model_model_language_model_layers_18_linear_attn_a_log": 241, + "p_model_model_language_model_layers_18_linear_attn_dt_bias": 242, + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight": 243, + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight": 244, + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight": 245, + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight": 246, + "p_model_model_language_model_layers_18_linear_attn_conv1d_weight": 247, + "p_model_model_language_model_layers_18_linear_attn_norm_weight": 248, + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight": 249, + "p_model_model_language_model_layers_18_post_attention_layernorm_weight": 250, + "p_model_model_language_model_layers_18_mlp_gate_proj_weight": 251, + "p_model_model_language_model_layers_18_mlp_up_proj_weight": 252, + "p_model_model_language_model_layers_18_mlp_down_proj_weight": 253, + "p_model_model_language_model_layers_19_input_layernorm_weight": 254, + "p_model_model_language_model_layers_19_self_attn_q_proj_weight": 255, + "p_model_model_language_model_layers_19_self_attn_q_norm_weight": 256, + "p_model_model_language_model_layers_19_self_attn_k_proj_weight": 257, + "p_model_model_language_model_layers_19_self_attn_k_norm_weight": 258, + "p_model_model_language_model_layers_19_self_attn_v_proj_weight": 259, + "p_model_model_language_model_layers_19_self_attn_o_proj_weight": 260, + "p_model_model_language_model_layers_19_post_attention_layernorm_weight": 261, + "p_model_model_language_model_layers_19_mlp_gate_proj_weight": 262, + "p_model_model_language_model_layers_19_mlp_up_proj_weight": 263, + "p_model_model_language_model_layers_19_mlp_down_proj_weight": 264, + "p_model_model_language_model_layers_20_input_layernorm_weight": 265, + "p_model_model_language_model_layers_20_linear_attn_a_log": 266, + "p_model_model_language_model_layers_20_linear_attn_dt_bias": 267, + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight": 268, + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight": 269, + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight": 270, + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight": 271, + "p_model_model_language_model_layers_20_linear_attn_conv1d_weight": 272, + "p_model_model_language_model_layers_20_linear_attn_norm_weight": 273, + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight": 274, + "p_model_model_language_model_layers_20_post_attention_layernorm_weight": 275, + "p_model_model_language_model_layers_20_mlp_gate_proj_weight": 276, + "p_model_model_language_model_layers_20_mlp_up_proj_weight": 277, + "p_model_model_language_model_layers_20_mlp_down_proj_weight": 278, + "p_model_model_language_model_layers_21_input_layernorm_weight": 279, + "p_model_model_language_model_layers_21_linear_attn_a_log": 280, + "p_model_model_language_model_layers_21_linear_attn_dt_bias": 281, + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight": 282, + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight": 283, + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight": 284, + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight": 285, + "p_model_model_language_model_layers_21_linear_attn_conv1d_weight": 286, + "p_model_model_language_model_layers_21_linear_attn_norm_weight": 287, + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight": 288, + "p_model_model_language_model_layers_21_post_attention_layernorm_weight": 289, + "p_model_model_language_model_layers_21_mlp_gate_proj_weight": 290, + "p_model_model_language_model_layers_21_mlp_up_proj_weight": 291, + "p_model_model_language_model_layers_21_mlp_down_proj_weight": 292, + "p_model_model_language_model_layers_22_input_layernorm_weight": 293, + "p_model_model_language_model_layers_22_linear_attn_a_log": 294, + "p_model_model_language_model_layers_22_linear_attn_dt_bias": 295, + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight": 296, + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight": 297, + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight": 298, + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight": 299, + "p_model_model_language_model_layers_22_linear_attn_conv1d_weight": 300, + "p_model_model_language_model_layers_22_linear_attn_norm_weight": 301, + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight": 302, + "p_model_model_language_model_layers_22_post_attention_layernorm_weight": 303, + "p_model_model_language_model_layers_22_mlp_gate_proj_weight": 304, + "p_model_model_language_model_layers_22_mlp_up_proj_weight": 305, + "p_model_model_language_model_layers_22_mlp_down_proj_weight": 306, + "p_model_model_language_model_layers_23_input_layernorm_weight": 307, + "p_model_model_language_model_layers_23_self_attn_q_proj_weight": 308, + "p_model_model_language_model_layers_23_self_attn_q_norm_weight": 309, + "p_model_model_language_model_layers_23_self_attn_k_proj_weight": 310, + "p_model_model_language_model_layers_23_self_attn_k_norm_weight": 311, + "p_model_model_language_model_layers_23_self_attn_v_proj_weight": 312, + "p_model_model_language_model_layers_23_self_attn_o_proj_weight": 313, + "p_model_model_language_model_layers_23_post_attention_layernorm_weight": 314, + "p_model_model_language_model_layers_23_mlp_gate_proj_weight": 315, + "p_model_model_language_model_layers_23_mlp_up_proj_weight": 316, + "p_model_model_language_model_layers_23_mlp_down_proj_weight": 317, + "p_model_model_language_model_norm_weight": 318, + "b_model_model_language_model_rotary_emb_inv_freq": 319, + "input_ids": 320, + "position_ids": 321, + "attention_mask": 322, + "states_0": 323, + "states_1": 324, + "states_2": 325, + "states_3": 326, + "states_4": 327, + "states_5": 328, + "states_6": 329, + "states_7": 330, + "states_8": 331, + "states_9": 332, + "states_10": 333, + "states_11": 334, + "states_12": 335, + "states_13": 336, + "states_14": 337, + "states_15": 338, + "states_16": 339, + "states_17": 340, + "states_18": 341, + "states_19": 342, + "states_20": 343, + "states_21": 344, + "states_22": 345, + "states_23": 346, + "states_24": 347, + "states_25": 348, + "states_26": 349, + "states_27": 350, + "states_28": 351, + "states_29": 352, + "states_30": 353, + "states_31": 354, + "states_32": 355, + "states_33": 356, + "states_34": 357, + "states_35": 358, + "states_36": 359, + "states_37": 360, + "states_38": 361, + "states_39": 362, + "states_40": 363, + "states_41": 364, + "states_42": 365, + "states_43": 366, + "states_44": 367, + "states_45": 368, + "states_46": 369, + "states_47": 370, + "p_model_model_language_model_layers_0_mlp_down_proj_weight__int4_scales": 371, + "p_model_model_language_model_layers_0_mlp_gate_proj_weight__int4_scales": 372, + "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales": 373, + "p_model_model_language_model_layers_10_mlp_down_proj_weight__int4_scales": 374, + "p_model_model_language_model_layers_10_mlp_gate_proj_weight__int4_scales": 375, + "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales": 376, + "p_model_model_language_model_layers_11_mlp_down_proj_weight__int4_scales": 377, + "p_model_model_language_model_layers_11_mlp_gate_proj_weight__int4_scales": 378, + "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales": 379, + "p_model_model_language_model_layers_12_mlp_down_proj_weight__int4_scales": 380, + "p_model_model_language_model_layers_12_mlp_gate_proj_weight__int4_scales": 381, + "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales": 382, + "p_model_model_language_model_layers_13_mlp_down_proj_weight__int4_scales": 383, + "p_model_model_language_model_layers_13_mlp_gate_proj_weight__int4_scales": 384, + "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales": 385, + "p_model_model_language_model_layers_14_mlp_down_proj_weight__int4_scales": 386, + "p_model_model_language_model_layers_14_mlp_gate_proj_weight__int4_scales": 387, + "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales": 388, + "p_model_model_language_model_layers_15_mlp_down_proj_weight__int4_scales": 389, + "p_model_model_language_model_layers_15_mlp_gate_proj_weight__int4_scales": 390, + "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales": 391, + "p_model_model_language_model_layers_16_mlp_down_proj_weight__int4_scales": 392, + "p_model_model_language_model_layers_16_mlp_gate_proj_weight__int4_scales": 393, + "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales": 394, + "p_model_model_language_model_layers_17_mlp_down_proj_weight__int4_scales": 395, + "p_model_model_language_model_layers_17_mlp_gate_proj_weight__int4_scales": 396, + "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales": 397, + "p_model_model_language_model_layers_18_mlp_down_proj_weight__int4_scales": 398, + "p_model_model_language_model_layers_18_mlp_gate_proj_weight__int4_scales": 399, + "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales": 400, + "p_model_model_language_model_layers_19_mlp_down_proj_weight__int4_scales": 401, + "p_model_model_language_model_layers_19_mlp_gate_proj_weight__int4_scales": 402, + "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales": 403, + "p_model_model_language_model_layers_1_mlp_down_proj_weight__int4_scales": 404, + "p_model_model_language_model_layers_1_mlp_gate_proj_weight__int4_scales": 405, + "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales": 406, + "p_model_model_language_model_layers_20_mlp_down_proj_weight__int4_scales": 407, + "p_model_model_language_model_layers_20_mlp_gate_proj_weight__int4_scales": 408, + "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales": 409, + "p_model_model_language_model_layers_21_mlp_down_proj_weight__int4_scales": 410, + "p_model_model_language_model_layers_21_mlp_gate_proj_weight__int4_scales": 411, + "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales": 412, + "p_model_model_language_model_layers_22_mlp_down_proj_weight__int4_scales": 413, + "p_model_model_language_model_layers_22_mlp_gate_proj_weight__int4_scales": 414, + "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales": 415, + "p_model_model_language_model_layers_23_mlp_down_proj_weight__int4_scales": 416, + "p_model_model_language_model_layers_23_mlp_gate_proj_weight__int4_scales": 417, + "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales": 418, + "p_model_model_language_model_layers_2_mlp_down_proj_weight__int4_scales": 419, + "p_model_model_language_model_layers_2_mlp_gate_proj_weight__int4_scales": 420, + "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales": 421, + "p_model_model_language_model_layers_3_mlp_down_proj_weight__int4_scales": 422, + "p_model_model_language_model_layers_3_mlp_gate_proj_weight__int4_scales": 423, + "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales": 424, + "p_model_model_language_model_layers_4_mlp_down_proj_weight__int4_scales": 425, + "p_model_model_language_model_layers_4_mlp_gate_proj_weight__int4_scales": 426, + "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales": 427, + "p_model_model_language_model_layers_5_mlp_down_proj_weight__int4_scales": 428, + "p_model_model_language_model_layers_5_mlp_gate_proj_weight__int4_scales": 429, + "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales": 430, + "p_model_model_language_model_layers_6_mlp_down_proj_weight__int4_scales": 431, + "p_model_model_language_model_layers_6_mlp_gate_proj_weight__int4_scales": 432, + "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales": 433, + "p_model_model_language_model_layers_7_mlp_down_proj_weight__int4_scales": 434, + "p_model_model_language_model_layers_7_mlp_gate_proj_weight__int4_scales": 435, + "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales": 436, + "p_model_model_language_model_layers_8_mlp_down_proj_weight__int4_scales": 437, + "p_model_model_language_model_layers_8_mlp_gate_proj_weight__int4_scales": 438, + "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales": 439, + "p_model_model_language_model_layers_9_mlp_down_proj_weight__int4_scales": 440, + "p_model_model_language_model_layers_9_mlp_gate_proj_weight__int4_scales": 441, + "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales": 442, + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight__int8_scales": 443, + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight__int8_scales": 444, + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight__int8_scales": 445, + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight__int8_scales": 446, + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight__int8_scales": 447, + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight__int8_scales": 448, + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight__int8_scales": 449, + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight__int8_scales": 450, + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight__int8_scales": 451, + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight__int8_scales": 452, + "p_model_model_language_model_layers_11_self_attn_k_proj_weight__int8_scales": 453, + "p_model_model_language_model_layers_11_self_attn_o_proj_weight__int8_scales": 454, + "p_model_model_language_model_layers_11_self_attn_q_proj_weight__int8_scales": 455, + "p_model_model_language_model_layers_11_self_attn_v_proj_weight__int8_scales": 456, + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight__int8_scales": 457, + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight__int8_scales": 458, + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight__int8_scales": 459, + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight__int8_scales": 460, + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight__int8_scales": 461, + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight__int8_scales": 462, + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight__int8_scales": 463, + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight__int8_scales": 464, + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight__int8_scales": 465, + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight__int8_scales": 466, + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight__int8_scales": 467, + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight__int8_scales": 468, + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight__int8_scales": 469, + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight__int8_scales": 470, + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight__int8_scales": 471, + "p_model_model_language_model_layers_15_self_attn_k_proj_weight__int8_scales": 472, + "p_model_model_language_model_layers_15_self_attn_o_proj_weight__int8_scales": 473, + "p_model_model_language_model_layers_15_self_attn_q_proj_weight__int8_scales": 474, + "p_model_model_language_model_layers_15_self_attn_v_proj_weight__int8_scales": 475, + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight__int8_scales": 476, + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight__int8_scales": 477, + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight__int8_scales": 478, + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight__int8_scales": 479, + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight__int8_scales": 480, + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight__int8_scales": 481, + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight__int8_scales": 482, + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight__int8_scales": 483, + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight__int8_scales": 484, + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight__int8_scales": 485, + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight__int8_scales": 486, + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight__int8_scales": 487, + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight__int8_scales": 488, + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight__int8_scales": 489, + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight__int8_scales": 490, + "p_model_model_language_model_layers_19_self_attn_k_proj_weight__int8_scales": 491, + "p_model_model_language_model_layers_19_self_attn_o_proj_weight__int8_scales": 492, + "p_model_model_language_model_layers_19_self_attn_q_proj_weight__int8_scales": 493, + "p_model_model_language_model_layers_19_self_attn_v_proj_weight__int8_scales": 494, + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight__int8_scales": 495, + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight__int8_scales": 496, + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight__int8_scales": 497, + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight__int8_scales": 498, + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight__int8_scales": 499, + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight__int8_scales": 500, + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight__int8_scales": 501, + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight__int8_scales": 502, + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight__int8_scales": 503, + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight__int8_scales": 504, + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight__int8_scales": 505, + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight__int8_scales": 506, + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight__int8_scales": 507, + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight__int8_scales": 508, + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight__int8_scales": 509, + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight__int8_scales": 510, + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight__int8_scales": 511, + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight__int8_scales": 512, + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight__int8_scales": 513, + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight__int8_scales": 514, + "p_model_model_language_model_layers_23_self_attn_k_proj_weight__int8_scales": 515, + "p_model_model_language_model_layers_23_self_attn_o_proj_weight__int8_scales": 516, + "p_model_model_language_model_layers_23_self_attn_q_proj_weight__int8_scales": 517, + "p_model_model_language_model_layers_23_self_attn_v_proj_weight__int8_scales": 518, + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight__int8_scales": 519, + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight__int8_scales": 520, + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight__int8_scales": 521, + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight__int8_scales": 522, + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight__int8_scales": 523, + "p_model_model_language_model_layers_3_self_attn_k_proj_weight__int8_scales": 524, + "p_model_model_language_model_layers_3_self_attn_o_proj_weight__int8_scales": 525, + "p_model_model_language_model_layers_3_self_attn_q_proj_weight__int8_scales": 526, + "p_model_model_language_model_layers_3_self_attn_v_proj_weight__int8_scales": 527, + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight__int8_scales": 528, + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight__int8_scales": 529, + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight__int8_scales": 530, + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight__int8_scales": 531, + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight__int8_scales": 532, + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight__int8_scales": 533, + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight__int8_scales": 534, + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight__int8_scales": 535, + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight__int8_scales": 536, + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight__int8_scales": 537, + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight__int8_scales": 538, + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight__int8_scales": 539, + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight__int8_scales": 540, + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight__int8_scales": 541, + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight__int8_scales": 542, + "p_model_model_language_model_layers_7_self_attn_k_proj_weight__int8_scales": 543, + "p_model_model_language_model_layers_7_self_attn_o_proj_weight__int8_scales": 544, + "p_model_model_language_model_layers_7_self_attn_q_proj_weight__int8_scales": 545, + "p_model_model_language_model_layers_7_self_attn_v_proj_weight__int8_scales": 546, + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight__int8_scales": 547, + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight__int8_scales": 548, + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight__int8_scales": 549, + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight__int8_scales": 550, + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight__int8_scales": 551, + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight__int8_scales": 552, + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight__int8_scales": 553, + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight__int8_scales": 554, + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight__int8_scales": 555, + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight__int8_scales": 556, + "p_model_model_language_model_embed_tokens_weight__rows_0_131072": 557, + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072": 558, + "p_model_model_language_model_embed_tokens_weight__rows_131072_117248": 559, + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248": 560, + "embedding__part_0": 561, + "embedding__part_1": 562, + "embedding": 563, + "unsqueeze": 561, + "expand": 562, + "slice_1": 561, + "expand_1": 562, + "_to_copy": 564, + "matmul": 561, + "transpose": 562, + "select": 565, + "select_1": 566, + "slice_2": 562, + "slice_3": 566, + "copy": 565, + "transpose_1": 562, + "select_2": 561, + "slice_scatter": 567, + "select_scatter": 561, + "transpose_2": 562, + "transpose_7": 561, + "select_7": 565, + "slice_7": 567, + "transpose_8": 561, + "select_8": 565, + "slice_8": 561, + "copy_1": 565, + "transpose_9": 561, + "select_9": 562, + "slice_scatter_1": 568, + "select_scatter_1": 562, + "transpose_10": 561, + "transpose_12": 562, + "select_11": 565, + "cat": 561, + "cos": 562, + "mul": 569, + "sin": 562, + "mul_1": 561, + "mul_3": 562, + "linear": 570, + "transpose_13": 571, + "linear_1": 570, + "linear_2": 568, + "linear_3": 565, + "cat_1": 572, + "slice_10": 323, + "conv1d": 573, + "slice_11": 572, + "silu": 573, + "slice_12": 571, + "transpose_14": 572, + "slice_13": 571, + "slice_14": 573, + "slice_15": 574, + "sigmoid": 562, + "exp": 568, + "neg": 572, + "add_2": 568, + "softplus": 565, + "mul_4": 568, + "mul_5": 572, + "sum_1": 565, + "add_3": 572, + "rsqrt_1": 565, + "mul_6": 572, + "mul_7": 571, + "sum_2": 565, + "add_4": 571, + "rsqrt_2": 565, + "mul_8": 571, + "normalized_gated_delta_rule": 575, + "slice_16": 324, + "slice_17": 574, + "transpose_15": 571, + "mul_10": 574, + "mul_11": 571, + "linear_4": 562, + "add_6": 574, + "mul_13": 563, + "linear_5": 570, + "linear_6": 571, + "mul_14": 572, + "linear_7": 562, + "add_9": 563, + "mul_16": 562, + "linear_8": 570, + "transpose_16": 571, + "linear_9": 574, + "linear_10": 565, + "linear_11": 568, + "cat_2": 572, + "slice_18": 325, + "conv1d_1": 573, + "slice_19": 572, + "silu_3": 573, + "slice_20": 570, + "transpose_17": 571, + "slice_21": 570, + "slice_22": 572, + "slice_23": 573, + "sigmoid_1": 562, + "exp_1": 565, + "neg_1": 571, + "add_12": 565, + "softplus_1": 568, + "mul_17": 565, + "mul_18": 571, + "sum_3": 568, + "add_13": 571, + "rsqrt_6": 568, + "mul_19": 571, + "mul_20": 570, + "sum_4": 568, + "add_14": 570, + "rsqrt_7": 568, + "mul_21": 570, + "normalized_gated_delta_rule_1": 575, + "slice_24": 326, + "slice_25": 571, + "transpose_18": 570, + "mul_23": 571, + "mul_24": 570, + "linear_12": 562, + "add_16": 574, + "mul_26": 563, + "linear_13": 571, + "linear_14": 570, + "mul_27": 572, + "linear_15": 562, + "add_19": 563, + "mul_29": 562, + "linear_16": 571, + "transpose_19": 570, + "linear_17": 574, + "linear_18": 568, + "linear_19": 565, + "cat_3": 572, + "slice_26": 327, + "conv1d_2": 573, + "slice_27": 572, + "silu_6": 573, + "slice_28": 571, + "transpose_20": 570, + "slice_29": 571, + "slice_30": 572, + "slice_31": 573, + "sigmoid_2": 562, + "exp_2": 568, + "neg_2": 570, + "add_22": 568, + "softplus_2": 565, + "mul_30": 568, + "mul_31": 570, + "sum_5": 565, + "add_23": 570, + "rsqrt_11": 565, + "mul_32": 570, + "mul_33": 571, + "sum_6": 565, + "add_24": 571, + "rsqrt_12": 565, + "mul_34": 571, + "normalized_gated_delta_rule_2": 575, + "slice_32": 328, + "slice_33": 570, + "transpose_21": 571, + "mul_36": 570, + "mul_37": 571, + "linear_20": 562, + "add_26": 574, + "mul_39": 563, + "linear_21": 570, + "linear_22": 571, + "mul_40": 572, + "linear_23": 562, + "add_29": 563, + "mul_42": 562, + "linear_24": 570, + "slice_34": 574, + "slice_35": 571, + "mul_44": 570, + "transpose_22": 574, + "linear_25": 570, + "mul_46": 572, + "transpose_23": 570, + "linear_26": 572, + "transpose_24": 562, + "slice_36": 572, + "slice_37": 573, + "slice_38": 574, + "slice_39": 575, + "mul_47": 570, + "slice_40": 576, + "slice_41": 577, + "neg_3": 572, + "cat_4": 578, + "mul_48": 572, + "add_36": 578, + "mul_49": 577, + "slice_42": 576, + "slice_43": 570, + "neg_4": 574, + "cat_5": 570, + "mul_50": 576, + "add_37": 574, + "cat_6": 570, + "cat_7": 578, + "cache_update": 329, + "cache_update_1": 330, + "matmul_2": 574, + "transpose_26": 570, + "sigmoid_3": 574, + "mul_52": 571, + "linear_27": 562, + "add_39": 574, + "mul_54": 563, + "linear_28": 570, + "linear_29": 571, + "mul_55": 572, + "linear_30": 562, + "add_42": 563, + "mul_57": 562, + "linear_31": 570, + "transpose_27": 571, + "linear_32": 574, + "linear_33": 565, + "linear_34": 568, + "cat_8": 572, + "slice_44": 331, + "conv1d_3": 573, + "slice_45": 572, + "silu_10": 573, + "slice_46": 570, + "transpose_28": 571, + "slice_47": 570, + "slice_48": 572, + "slice_49": 573, + "sigmoid_4": 577, + "exp_3": 565, + "neg_5": 576, + "add_45": 565, + "softplus_3": 568, + "mul_58": 565, + "mul_59": 571, + "sum_7": 568, + "add_46": 576, + "rsqrt_20": 568, + "mul_60": 571, + "mul_61": 570, + "sum_8": 568, + "add_47": 576, + "rsqrt_21": 568, + "mul_62": 570, + "normalized_gated_delta_rule_3": 575, + "slice_50": 332, + "slice_51": 571, + "transpose_29": 570, + "mul_64": 571, + "mul_65": 570, + "linear_35": 562, + "add_49": 574, + "mul_67": 563, + "linear_36": 571, + "linear_37": 570, + "mul_68": 572, + "linear_38": 562, + "add_52": 563, + "mul_70": 562, + "linear_39": 571, + "transpose_30": 570, + "linear_40": 574, + "linear_41": 568, + "linear_42": 565, + "cat_9": 572, + "slice_52": 333, + "conv1d_4": 573, + "slice_53": 572, + "silu_13": 573, + "slice_54": 571, + "transpose_31": 570, + "slice_55": 571, + "slice_56": 572, + "slice_57": 573, + "sigmoid_5": 576, + "exp_4": 568, + "neg_6": 577, + "add_55": 568, + "softplus_4": 565, + "mul_71": 568, + "mul_72": 570, + "sum_9": 565, + "add_56": 577, + "rsqrt_25": 565, + "mul_73": 570, + "mul_74": 571, + "sum_10": 565, + "add_57": 577, + "rsqrt_26": 565, + "mul_75": 571, + "normalized_gated_delta_rule_4": 575, + "slice_58": 334, + "slice_59": 570, + "transpose_32": 571, + "mul_77": 570, + "mul_78": 571, + "linear_43": 562, + "add_59": 574, + "mul_80": 563, + "linear_44": 570, + "linear_45": 571, + "mul_81": 572, + "linear_46": 562, + "add_62": 563, + "mul_83": 562, + "linear_47": 570, + "transpose_33": 571, + "linear_48": 574, + "linear_49": 565, + "linear_50": 568, + "cat_10": 572, + "slice_60": 335, + "conv1d_5": 573, + "slice_61": 572, + "silu_16": 573, + "slice_62": 570, + "transpose_34": 571, + "slice_63": 570, + "slice_64": 572, + "slice_65": 573, + "sigmoid_6": 577, + "exp_5": 565, + "neg_7": 576, + "add_65": 565, + "softplus_5": 568, + "mul_84": 565, + "mul_85": 571, + "sum_11": 568, + "add_66": 576, + "rsqrt_30": 568, + "mul_86": 571, + "mul_87": 570, + "sum_12": 568, + "add_67": 576, + "rsqrt_31": 568, + "mul_88": 570, + "normalized_gated_delta_rule_5": 575, + "slice_66": 336, + "slice_67": 571, + "transpose_35": 570, + "mul_90": 571, + "mul_91": 570, + "linear_51": 562, + "add_69": 574, + "mul_93": 563, + "linear_52": 571, + "linear_53": 570, + "mul_94": 572, + "linear_54": 562, + "add_72": 563, + "mul_96": 562, + "linear_55": 571, + "slice_68": 574, + "slice_69": 570, + "mul_98": 571, + "transpose_36": 574, + "linear_56": 578, + "mul_100": 571, + "transpose_37": 578, + "linear_57": 571, + "transpose_38": 562, + "slice_70": 571, + "slice_71": 572, + "slice_72": 576, + "slice_73": 574, + "mul_101": 578, + "slice_74": 577, + "slice_75": 573, + "neg_8": 571, + "cat_11": 573, + "mul_102": 571, + "add_79": 573, + "mul_103": 577, + "slice_76": 578, + "slice_77": 571, + "neg_9": 576, + "cat_12": 571, + "mul_104": 576, + "add_80": 578, + "cat_13": 571, + "cat_14": 572, + "cache_update_2": 337, + "cache_update_3": 338, + "matmul_4": 574, + "transpose_40": 571, + "sigmoid_7": 574, + "mul_106": 570, + "linear_58": 562, + "add_82": 574, + "mul_108": 563, + "linear_59": 571, + "linear_60": 570, + "mul_109": 572, + "linear_61": 562, + "add_85": 563, + "mul_111": 562, + "linear_62": 571, + "transpose_41": 570, + "linear_63": 574, + "linear_64": 568, + "linear_65": 565, + "cat_15": 572, + "slice_78": 339, + "conv1d_6": 573, + "slice_79": 572, + "silu_20": 573, + "slice_80": 571, + "transpose_42": 570, + "slice_81": 571, + "slice_82": 572, + "slice_83": 573, + "sigmoid_8": 577, + "exp_6": 568, + "neg_10": 576, + "add_88": 568, + "softplus_6": 565, + "mul_112": 568, + "mul_113": 570, + "sum_13": 565, + "add_89": 576, + "rsqrt_39": 565, + "mul_114": 570, + "mul_115": 571, + "sum_14": 565, + "add_90": 576, + "rsqrt_40": 565, + "mul_116": 571, + "normalized_gated_delta_rule_6": 575, + "slice_84": 340, + "slice_85": 570, + "transpose_43": 571, + "mul_118": 570, + "mul_119": 571, + "linear_66": 562, + "add_92": 574, + "mul_121": 563, + "linear_67": 570, + "linear_68": 571, + "mul_122": 572, + "linear_69": 562, + "add_95": 563, + "mul_124": 562, + "linear_70": 570, + "transpose_44": 571, + "linear_71": 574, + "linear_72": 565, + "linear_73": 568, + "cat_16": 572, + "slice_86": 341, + "conv1d_7": 573, + "slice_87": 572, + "silu_23": 573, + "slice_88": 570, + "transpose_45": 571, + "slice_89": 570, + "slice_90": 572, + "slice_91": 573, + "sigmoid_9": 576, + "exp_7": 565, + "neg_11": 577, + "add_98": 565, + "softplus_7": 568, + "mul_125": 565, + "mul_126": 571, + "sum_15": 568, + "add_99": 577, + "rsqrt_44": 568, + "mul_127": 571, + "mul_128": 570, + "sum_16": 568, + "add_100": 577, + "rsqrt_45": 568, + "mul_129": 570, + "normalized_gated_delta_rule_7": 575, + "slice_92": 342, + "slice_93": 571, + "transpose_46": 570, + "mul_131": 571, + "mul_132": 570, + "linear_74": 562, + "add_102": 574, + "mul_134": 563, + "linear_75": 571, + "linear_76": 570, + "mul_135": 572, + "linear_77": 562, + "add_105": 563, + "mul_137": 562, + "linear_78": 571, + "transpose_47": 570, + "linear_79": 574, + "linear_80": 568, + "linear_81": 565, + "cat_17": 572, + "slice_94": 343, + "conv1d_8": 573, + "slice_95": 572, + "silu_26": 573, + "slice_96": 571, + "transpose_48": 570, + "slice_97": 571, + "slice_98": 572, + "slice_99": 573, + "sigmoid_10": 577, + "exp_8": 568, + "neg_12": 576, + "add_108": 568, + "softplus_8": 565, + "mul_138": 568, + "mul_139": 570, + "sum_17": 565, + "add_109": 576, + "rsqrt_49": 565, + "mul_140": 570, + "mul_141": 571, + "sum_18": 565, + "add_110": 576, + "rsqrt_50": 565, + "mul_142": 571, + "normalized_gated_delta_rule_8": 575, + "slice_100": 344, + "slice_101": 570, + "transpose_49": 571, + "mul_144": 570, + "mul_145": 571, + "linear_82": 562, + "add_112": 574, + "mul_147": 563, + "linear_83": 570, + "linear_84": 571, + "mul_148": 572, + "linear_85": 562, + "add_115": 563, + "mul_150": 562, + "linear_86": 570, + "slice_102": 574, + "slice_103": 571, + "mul_152": 570, + "transpose_50": 574, + "linear_87": 578, + "mul_154": 570, + "transpose_51": 578, + "linear_88": 570, + "transpose_52": 562, + "slice_104": 570, + "slice_105": 572, + "slice_106": 576, + "slice_107": 574, + "mul_155": 578, + "slice_108": 577, + "slice_109": 573, + "neg_13": 570, + "cat_18": 573, + "mul_156": 570, + "add_122": 573, + "mul_157": 577, + "slice_110": 578, + "slice_111": 570, + "neg_14": 576, + "cat_19": 570, + "mul_158": 576, + "add_123": 578, + "cat_20": 570, + "cat_21": 572, + "cache_update_4": 345, + "cache_update_5": 346, + "matmul_6": 574, + "transpose_54": 570, + "sigmoid_11": 574, + "mul_160": 571, + "linear_89": 562, + "add_125": 574, + "mul_162": 563, + "linear_90": 570, + "linear_91": 571, + "mul_163": 572, + "linear_92": 562, + "add_128": 563, + "mul_165": 562, + "linear_93": 570, + "transpose_55": 571, + "linear_94": 574, + "linear_95": 565, + "linear_96": 568, + "cat_22": 572, + "slice_112": 347, + "conv1d_9": 573, + "slice_113": 572, + "silu_30": 573, + "slice_114": 570, + "transpose_56": 571, + "slice_115": 570, + "slice_116": 572, + "slice_117": 573, + "sigmoid_12": 577, + "exp_9": 565, + "neg_15": 576, + "add_131": 565, + "softplus_9": 568, + "mul_166": 565, + "mul_167": 571, + "sum_19": 568, + "add_132": 576, + "rsqrt_58": 568, + "mul_168": 571, + "mul_169": 570, + "sum_20": 568, + "add_133": 576, + "rsqrt_59": 568, + "mul_170": 570, + "normalized_gated_delta_rule_9": 575, + "slice_118": 348, + "slice_119": 571, + "transpose_57": 570, + "mul_172": 571, + "mul_173": 570, + "linear_97": 562, + "add_135": 574, + "mul_175": 563, + "linear_98": 571, + "linear_99": 570, + "mul_176": 572, + "linear_100": 562, + "add_138": 563, + "mul_178": 562, + "linear_101": 571, + "transpose_58": 570, + "linear_102": 574, + "linear_103": 568, + "linear_104": 565, + "cat_23": 572, + "slice_120": 349, + "conv1d_10": 573, + "slice_121": 572, + "silu_33": 573, + "slice_122": 571, + "transpose_59": 570, + "slice_123": 571, + "slice_124": 572, + "slice_125": 573, + "sigmoid_13": 576, + "exp_10": 568, + "neg_16": 577, + "add_141": 568, + "softplus_10": 565, + "mul_179": 568, + "mul_180": 570, + "sum_21": 565, + "add_142": 577, + "rsqrt_63": 565, + "mul_181": 570, + "mul_182": 571, + "sum_22": 565, + "add_143": 577, + "rsqrt_64": 565, + "mul_183": 571, + "normalized_gated_delta_rule_10": 575, + "slice_126": 350, + "slice_127": 570, + "transpose_60": 571, + "mul_185": 570, + "mul_186": 571, + "linear_105": 562, + "add_145": 574, + "mul_188": 563, + "linear_106": 570, + "linear_107": 571, + "mul_189": 572, + "linear_108": 562, + "add_148": 563, + "mul_191": 562, + "linear_109": 570, + "transpose_61": 571, + "linear_110": 574, + "linear_111": 565, + "linear_112": 568, + "cat_24": 572, + "slice_128": 351, + "conv1d_11": 573, + "slice_129": 572, + "silu_36": 573, + "slice_130": 570, + "transpose_62": 571, + "slice_131": 570, + "slice_132": 572, + "slice_133": 573, + "sigmoid_14": 577, + "exp_11": 565, + "neg_17": 576, + "add_151": 565, + "softplus_11": 568, + "mul_192": 565, + "mul_193": 571, + "sum_23": 568, + "add_152": 576, + "rsqrt_68": 568, + "mul_194": 571, + "mul_195": 570, + "sum_24": 568, + "add_153": 576, + "rsqrt_69": 568, + "mul_196": 570, + "normalized_gated_delta_rule_11": 575, + "slice_134": 352, + "slice_135": 571, + "transpose_63": 570, + "mul_198": 571, + "mul_199": 570, + "linear_113": 562, + "add_155": 574, + "mul_201": 563, + "linear_114": 571, + "linear_115": 570, + "mul_202": 572, + "linear_116": 562, + "add_158": 563, + "mul_204": 562, + "linear_117": 571, + "slice_136": 574, + "slice_137": 570, + "mul_206": 571, + "transpose_64": 574, + "linear_118": 578, + "mul_208": 571, + "transpose_65": 578, + "linear_119": 571, + "transpose_66": 562, + "slice_138": 571, + "slice_139": 572, + "slice_140": 576, + "slice_141": 574, + "mul_209": 578, + "slice_142": 577, + "slice_143": 573, + "neg_18": 571, + "cat_25": 573, + "mul_210": 571, + "add_165": 573, + "mul_211": 577, + "slice_144": 578, + "slice_145": 571, + "neg_19": 576, + "cat_26": 571, + "mul_212": 576, + "add_166": 578, + "cat_27": 571, + "cat_28": 572, + "cache_update_6": 353, + "cache_update_7": 354, + "matmul_8": 574, + "transpose_68": 571, + "sigmoid_15": 574, + "mul_214": 570, + "linear_120": 562, + "add_168": 574, + "mul_216": 563, + "linear_121": 571, + "linear_122": 570, + "mul_217": 572, + "linear_123": 562, + "add_171": 563, + "mul_219": 562, + "linear_124": 571, + "transpose_69": 570, + "linear_125": 574, + "linear_126": 568, + "linear_127": 565, + "cat_29": 572, + "slice_146": 355, + "conv1d_12": 573, + "slice_147": 572, + "silu_40": 573, + "slice_148": 571, + "transpose_70": 570, + "slice_149": 571, + "slice_150": 572, + "slice_151": 573, + "sigmoid_16": 577, + "exp_12": 568, + "neg_20": 576, + "add_174": 568, + "softplus_12": 565, + "mul_220": 568, + "mul_221": 570, + "sum_25": 565, + "add_175": 576, + "rsqrt_77": 565, + "mul_222": 570, + "mul_223": 571, + "sum_26": 565, + "add_176": 576, + "rsqrt_78": 565, + "mul_224": 571, + "normalized_gated_delta_rule_12": 575, + "slice_152": 356, + "slice_153": 570, + "transpose_71": 571, + "mul_226": 570, + "mul_227": 571, + "linear_128": 562, + "add_178": 574, + "mul_229": 563, + "linear_129": 570, + "linear_130": 571, + "mul_230": 572, + "linear_131": 562, + "add_181": 563, + "mul_232": 562, + "linear_132": 570, + "transpose_72": 571, + "linear_133": 574, + "linear_134": 565, + "linear_135": 568, + "cat_30": 572, + "slice_154": 357, + "conv1d_13": 573, + "slice_155": 572, + "silu_43": 573, + "slice_156": 570, + "transpose_73": 571, + "slice_157": 570, + "slice_158": 572, + "slice_159": 573, + "sigmoid_17": 576, + "exp_13": 565, + "neg_21": 577, + "add_184": 565, + "softplus_13": 568, + "mul_233": 565, + "mul_234": 571, + "sum_27": 568, + "add_185": 577, + "rsqrt_82": 568, + "mul_235": 571, + "mul_236": 570, + "sum_28": 568, + "add_186": 577, + "rsqrt_83": 568, + "mul_237": 570, + "normalized_gated_delta_rule_13": 575, + "slice_160": 358, + "slice_161": 571, + "transpose_74": 570, + "mul_239": 571, + "mul_240": 570, + "linear_136": 562, + "add_188": 574, + "mul_242": 563, + "linear_137": 571, + "linear_138": 570, + "mul_243": 572, + "linear_139": 562, + "add_191": 563, + "mul_245": 562, + "linear_140": 571, + "transpose_75": 570, + "linear_141": 574, + "linear_142": 568, + "linear_143": 565, + "cat_31": 572, + "slice_162": 359, + "conv1d_14": 573, + "slice_163": 572, + "silu_46": 573, + "slice_164": 571, + "transpose_76": 570, + "slice_165": 571, + "slice_166": 572, + "slice_167": 573, + "sigmoid_18": 577, + "exp_14": 568, + "neg_22": 576, + "add_194": 568, + "softplus_14": 565, + "mul_246": 568, + "mul_247": 570, + "sum_29": 565, + "add_195": 576, + "rsqrt_87": 565, + "mul_248": 570, + "mul_249": 571, + "sum_30": 565, + "add_196": 576, + "rsqrt_88": 565, + "mul_250": 571, + "normalized_gated_delta_rule_14": 575, + "slice_168": 360, + "slice_169": 570, + "transpose_77": 571, + "mul_252": 570, + "mul_253": 571, + "linear_144": 562, + "add_198": 574, + "mul_255": 563, + "linear_145": 570, + "linear_146": 571, + "mul_256": 572, + "linear_147": 562, + "add_201": 563, + "mul_258": 562, + "linear_148": 570, + "slice_170": 574, + "slice_171": 571, + "mul_260": 570, + "transpose_78": 574, + "linear_149": 578, + "mul_262": 570, + "transpose_79": 578, + "linear_150": 570, + "transpose_80": 562, + "slice_172": 570, + "slice_173": 572, + "slice_174": 576, + "slice_175": 574, + "mul_263": 578, + "slice_176": 577, + "slice_177": 573, + "neg_23": 570, + "cat_32": 573, + "mul_264": 570, + "add_208": 573, + "mul_265": 577, + "slice_178": 578, + "slice_179": 570, + "neg_24": 576, + "cat_33": 570, + "mul_266": 576, + "add_209": 578, + "cat_34": 570, + "cat_35": 572, + "cache_update_8": 361, + "cache_update_9": 362, + "matmul_10": 574, + "transpose_82": 570, + "sigmoid_19": 574, + "mul_268": 571, + "linear_151": 562, + "add_211": 574, + "mul_270": 563, + "linear_152": 570, + "linear_153": 571, + "mul_271": 572, + "linear_154": 562, + "add_214": 563, + "mul_273": 562, + "linear_155": 570, + "transpose_83": 571, + "linear_156": 574, + "linear_157": 565, + "linear_158": 568, + "cat_36": 572, + "slice_180": 363, + "conv1d_15": 573, + "slice_181": 572, + "silu_50": 573, + "slice_182": 570, + "transpose_84": 571, + "slice_183": 570, + "slice_184": 572, + "slice_185": 573, + "sigmoid_20": 577, + "exp_15": 565, + "neg_25": 576, + "add_217": 565, + "softplus_15": 568, + "mul_274": 565, + "mul_275": 571, + "sum_31": 568, + "add_218": 576, + "rsqrt_96": 568, + "mul_276": 571, + "mul_277": 570, + "sum_32": 568, + "add_219": 576, + "rsqrt_97": 568, + "mul_278": 570, + "normalized_gated_delta_rule_15": 575, + "slice_186": 364, + "slice_187": 571, + "transpose_85": 570, + "mul_280": 571, + "mul_281": 570, + "linear_159": 562, + "add_221": 574, + "mul_283": 563, + "linear_160": 571, + "linear_161": 570, + "mul_284": 572, + "linear_162": 562, + "add_224": 563, + "mul_286": 562, + "linear_163": 571, + "transpose_86": 570, + "linear_164": 574, + "linear_165": 568, + "linear_166": 565, + "cat_37": 572, + "slice_188": 365, + "conv1d_16": 573, + "slice_189": 572, + "silu_53": 573, + "slice_190": 571, + "transpose_87": 570, + "slice_191": 571, + "slice_192": 572, + "slice_193": 573, + "sigmoid_21": 576, + "exp_16": 568, + "neg_26": 577, + "add_227": 568, + "softplus_16": 565, + "mul_287": 568, + "mul_288": 570, + "sum_33": 565, + "add_228": 577, + "rsqrt_101": 565, + "mul_289": 570, + "mul_290": 571, + "sum_34": 565, + "add_229": 577, + "rsqrt_102": 565, + "mul_291": 571, + "normalized_gated_delta_rule_16": 575, + "slice_194": 366, + "slice_195": 570, + "transpose_88": 571, + "mul_293": 570, + "mul_294": 571, + "linear_167": 562, + "add_231": 574, + "mul_296": 563, + "linear_168": 570, + "linear_169": 571, + "mul_297": 572, + "linear_170": 562, + "add_234": 563, + "mul_299": 562, + "linear_171": 570, + "transpose_89": 571, + "linear_172": 574, + "linear_173": 565, + "linear_174": 568, + "cat_38": 572, + "slice_196": 367, + "conv1d_17": 573, + "slice_197": 572, + "silu_56": 573, + "slice_198": 570, + "transpose_90": 571, + "slice_199": 570, + "slice_200": 572, + "slice_201": 573, + "sigmoid_22": 577, + "exp_17": 565, + "neg_27": 576, + "add_237": 565, + "softplus_17": 568, + "mul_300": 565, + "mul_301": 571, + "sum_35": 568, + "add_238": 576, + "rsqrt_106": 568, + "mul_302": 571, + "mul_303": 570, + "sum_36": 568, + "add_239": 576, + "rsqrt_107": 568, + "mul_304": 570, + "normalized_gated_delta_rule_17": 575, + "slice_202": 368, + "slice_203": 571, + "transpose_91": 570, + "mul_306": 571, + "mul_307": 570, + "linear_175": 562, + "add_241": 574, + "mul_309": 563, + "linear_176": 571, + "linear_177": 570, + "mul_310": 572, + "linear_178": 562, + "add_244": 563, + "mul_312": 562, + "linear_179": 571, + "slice_204": 574, + "slice_205": 570, + "mul_314": 571, + "transpose_92": 574, + "linear_180": 578, + "mul_316": 571, + "transpose_93": 578, + "linear_181": 571, + "transpose_94": 562, + "slice_206": 571, + "slice_207": 572, + "slice_208": 576, + "slice_209": 574, + "mul_317": 578, + "slice_210": 577, + "slice_211": 573, + "neg_28": 571, + "cat_39": 573, + "mul_318": 571, + "add_251": 573, + "mul_319": 577, + "slice_212": 569, + "slice_213": 578, + "neg_29": 576, + "cat_40": 578, + "mul_320": 576, + "add_252": 578, + "cat_41": 571, + "cat_42": 561, + "cache_update_10": 369, + "cache_update_11": 370, + "matmul_12": 574, + "transpose_96": 571, + "sigmoid_23": 574, + "mul_322": 570, + "linear_182": 561, + "add_254": 562, + "mul_324": 563, + "linear_183": 571, + "linear_184": 570, + "mul_325": 572, + "linear_185": 561, + "add_257": 563, + "mul_327": 562, + "linear_186__part_0": 579, + "linear_186__part_1": 580, + "linear_186": 581 + }, + "byteLength": 846920144 + }, + "rewrites": [ + { + "pattern": "rms_norm", + "nodes": [ + "pow_1", + "mean", + "add", + "rsqrt", + "mul_2", + "add_1", + "mul_3" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_2", + "mean_1", + "add_5", + "rsqrt_3", + "mul_9", + "mul_10" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_3", + "mean_2", + "add_7", + "rsqrt_4", + "mul_12", + "add_8", + "mul_13" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_4", + "mean_3", + "add_10", + "rsqrt_5", + "mul_15", + "add_11", + "mul_16" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_5", + "mean_4", + "add_15", + "rsqrt_8", + "mul_22", + "mul_23" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_6", + "mean_5", + "add_17", + "rsqrt_9", + "mul_25", + "add_18", + "mul_26" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_7", + "mean_6", + "add_20", + "rsqrt_10", + "mul_28", + "add_21", + "mul_29" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_8", + "mean_7", + "add_25", + "rsqrt_13", + "mul_35", + "mul_36" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_9", + "mean_8", + "add_27", + "rsqrt_14", + "mul_38", + "add_28", + "mul_39" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_10", + "mean_9", + "add_30", + "rsqrt_15", + "mul_41", + "add_31", + "mul_42" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_11", + "mean_10", + "add_32", + "rsqrt_16", + "mul_43", + "add_33", + "mul_44" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_12", + "mean_11", + "add_34", + "rsqrt_17", + "mul_45", + "add_35", + "mul_46" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_13", + "mean_12", + "add_40", + "rsqrt_18", + "mul_53", + "add_41", + "mul_54" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_14", + "mean_13", + "add_43", + "rsqrt_19", + "mul_56", + "add_44", + "mul_57" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_15", + "mean_14", + "add_48", + "rsqrt_22", + "mul_63", + "mul_64" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_16", + "mean_15", + "add_50", + "rsqrt_23", + "mul_66", + "add_51", + "mul_67" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_17", + "mean_16", + "add_53", + "rsqrt_24", + "mul_69", + "add_54", + "mul_70" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_18", + "mean_17", + "add_58", + "rsqrt_27", + "mul_76", + "mul_77" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_19", + "mean_18", + "add_60", + "rsqrt_28", + "mul_79", + "add_61", + "mul_80" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_20", + "mean_19", + "add_63", + "rsqrt_29", + "mul_82", + "add_64", + "mul_83" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_21", + "mean_20", + "add_68", + "rsqrt_32", + "mul_89", + "mul_90" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_22", + "mean_21", + "add_70", + "rsqrt_33", + "mul_92", + "add_71", + "mul_93" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_23", + "mean_22", + "add_73", + "rsqrt_34", + "mul_95", + "add_74", + "mul_96" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_24", + "mean_23", + "add_75", + "rsqrt_35", + "mul_97", + "add_76", + "mul_98" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_25", + "mean_24", + "add_77", + "rsqrt_36", + "mul_99", + "add_78", + "mul_100" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_26", + "mean_25", + "add_83", + "rsqrt_37", + "mul_107", + "add_84", + "mul_108" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_27", + "mean_26", + "add_86", + "rsqrt_38", + "mul_110", + "add_87", + "mul_111" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_28", + "mean_27", + "add_91", + "rsqrt_41", + "mul_117", + "mul_118" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_29", + "mean_28", + "add_93", + "rsqrt_42", + "mul_120", + "add_94", + "mul_121" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_30", + "mean_29", + "add_96", + "rsqrt_43", + "mul_123", + "add_97", + "mul_124" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_31", + "mean_30", + "add_101", + "rsqrt_46", + "mul_130", + "mul_131" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_32", + "mean_31", + "add_103", + "rsqrt_47", + "mul_133", + "add_104", + "mul_134" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_33", + "mean_32", + "add_106", + "rsqrt_48", + "mul_136", + "add_107", + "mul_137" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_34", + "mean_33", + "add_111", + "rsqrt_51", + "mul_143", + "mul_144" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_35", + "mean_34", + "add_113", + "rsqrt_52", + "mul_146", + "add_114", + "mul_147" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_36", + "mean_35", + "add_116", + "rsqrt_53", + "mul_149", + "add_117", + "mul_150" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_37", + "mean_36", + "add_118", + "rsqrt_54", + "mul_151", + "add_119", + "mul_152" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_38", + "mean_37", + "add_120", + "rsqrt_55", + "mul_153", + "add_121", + "mul_154" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_39", + "mean_38", + "add_126", + "rsqrt_56", + "mul_161", + "add_127", + "mul_162" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_40", + "mean_39", + "add_129", + "rsqrt_57", + "mul_164", + "add_130", + "mul_165" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_41", + "mean_40", + "add_134", + "rsqrt_60", + "mul_171", + "mul_172" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_42", + "mean_41", + "add_136", + "rsqrt_61", + "mul_174", + "add_137", + "mul_175" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_43", + "mean_42", + "add_139", + "rsqrt_62", + "mul_177", + "add_140", + "mul_178" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_44", + "mean_43", + "add_144", + "rsqrt_65", + "mul_184", + "mul_185" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_45", + "mean_44", + "add_146", + "rsqrt_66", + "mul_187", + "add_147", + "mul_188" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_46", + "mean_45", + "add_149", + "rsqrt_67", + "mul_190", + "add_150", + "mul_191" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_47", + "mean_46", + "add_154", + "rsqrt_70", + "mul_197", + "mul_198" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_48", + "mean_47", + "add_156", + "rsqrt_71", + "mul_200", + "add_157", + "mul_201" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_49", + "mean_48", + "add_159", + "rsqrt_72", + "mul_203", + "add_160", + "mul_204" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_50", + "mean_49", + "add_161", + "rsqrt_73", + "mul_205", + "add_162", + "mul_206" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_51", + "mean_50", + "add_163", + "rsqrt_74", + "mul_207", + "add_164", + "mul_208" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_52", + "mean_51", + "add_169", + "rsqrt_75", + "mul_215", + "add_170", + "mul_216" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_53", + "mean_52", + "add_172", + "rsqrt_76", + "mul_218", + "add_173", + "mul_219" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_54", + "mean_53", + "add_177", + "rsqrt_79", + "mul_225", + "mul_226" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_55", + "mean_54", + "add_179", + "rsqrt_80", + "mul_228", + "add_180", + "mul_229" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_56", + "mean_55", + "add_182", + "rsqrt_81", + "mul_231", + "add_183", + "mul_232" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_57", + "mean_56", + "add_187", + "rsqrt_84", + "mul_238", + "mul_239" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_58", + "mean_57", + "add_189", + "rsqrt_85", + "mul_241", + "add_190", + "mul_242" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_59", + "mean_58", + "add_192", + "rsqrt_86", + "mul_244", + "add_193", + "mul_245" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_60", + "mean_59", + "add_197", + "rsqrt_89", + "mul_251", + "mul_252" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_61", + "mean_60", + "add_199", + "rsqrt_90", + "mul_254", + "add_200", + "mul_255" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_62", + "mean_61", + "add_202", + "rsqrt_91", + "mul_257", + "add_203", + "mul_258" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_63", + "mean_62", + "add_204", + "rsqrt_92", + "mul_259", + "add_205", + "mul_260" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_64", + "mean_63", + "add_206", + "rsqrt_93", + "mul_261", + "add_207", + "mul_262" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_65", + "mean_64", + "add_212", + "rsqrt_94", + "mul_269", + "add_213", + "mul_270" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_66", + "mean_65", + "add_215", + "rsqrt_95", + "mul_272", + "add_216", + "mul_273" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_67", + "mean_66", + "add_220", + "rsqrt_98", + "mul_279", + "mul_280" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_68", + "mean_67", + "add_222", + "rsqrt_99", + "mul_282", + "add_223", + "mul_283" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_69", + "mean_68", + "add_225", + "rsqrt_100", + "mul_285", + "add_226", + "mul_286" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_70", + "mean_69", + "add_230", + "rsqrt_103", + "mul_292", + "mul_293" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_71", + "mean_70", + "add_232", + "rsqrt_104", + "mul_295", + "add_233", + "mul_296" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_72", + "mean_71", + "add_235", + "rsqrt_105", + "mul_298", + "add_236", + "mul_299" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_73", + "mean_72", + "add_240", + "rsqrt_108", + "mul_305", + "mul_306" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_74", + "mean_73", + "add_242", + "rsqrt_109", + "mul_308", + "add_243", + "mul_309" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_75", + "mean_74", + "add_245", + "rsqrt_110", + "mul_311", + "add_246", + "mul_312" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_76", + "mean_75", + "add_247", + "rsqrt_111", + "mul_313", + "add_248", + "mul_314" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_77", + "mean_76", + "add_249", + "rsqrt_112", + "mul_315", + "add_250", + "mul_316" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_78", + "mean_77", + "add_255", + "rsqrt_113", + "mul_323", + "add_256", + "mul_324" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_79", + "mean_78", + "add_258", + "rsqrt_114", + "mul_326", + "add_259", + "mul_327" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_1", + "mul_11" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_2", + "mul_14" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_4", + "mul_24" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_5", + "mul_27" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_7", + "mul_37" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_8", + "mul_40" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_9", + "mul_55" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_11", + "mul_65" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_12", + "mul_68" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_14", + "mul_78" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_15", + "mul_81" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_17", + "mul_91" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_18", + "mul_94" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_19", + "mul_109" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_21", + "mul_119" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_22", + "mul_122" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_24", + "mul_132" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_25", + "mul_135" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_27", + "mul_145" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_28", + "mul_148" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_29", + "mul_163" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_31", + "mul_173" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_32", + "mul_176" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_34", + "mul_186" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_35", + "mul_189" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_37", + "mul_199" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_38", + "mul_202" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_39", + "mul_217" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_41", + "mul_227" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_42", + "mul_230" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_44", + "mul_240" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_45", + "mul_243" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_47", + "mul_253" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_48", + "mul_256" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_49", + "mul_271" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_51", + "mul_281" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_52", + "mul_284" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_54", + "mul_294" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_55", + "mul_297" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_57", + "mul_307" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_58", + "mul_310" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_59", + "mul_325" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_1", + "clone_4", + "expand_2", + "unsqueeze_10", + "_unsafe_view_2", + "clone_5", + "expand_3", + "unsqueeze_11", + "transpose_25", + "matmul_1", + "mul_51", + "add_38", + "softmax", + "matmul_2" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_4", + "clone_11", + "expand_4", + "unsqueeze_17", + "_unsafe_view_5", + "clone_12", + "expand_5", + "unsqueeze_18", + "transpose_39", + "matmul_3", + "mul_105", + "add_81", + "softmax_1", + "matmul_4" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_7", + "clone_18", + "expand_6", + "unsqueeze_24", + "_unsafe_view_8", + "clone_19", + "expand_7", + "unsqueeze_25", + "transpose_53", + "matmul_5", + "mul_159", + "add_124", + "softmax_2", + "matmul_6" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_10", + "clone_25", + "expand_8", + "unsqueeze_31", + "_unsafe_view_11", + "clone_26", + "expand_9", + "unsqueeze_32", + "transpose_67", + "matmul_7", + "mul_213", + "add_167", + "softmax_3", + "matmul_8" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_13", + "clone_32", + "expand_10", + "unsqueeze_38", + "_unsafe_view_14", + "clone_33", + "expand_11", + "unsqueeze_39", + "transpose_81", + "matmul_9", + "mul_267", + "add_210", + "softmax_4", + "matmul_10" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_16", + "clone_39", + "expand_12", + "unsqueeze_45", + "_unsafe_view_17", + "clone_40", + "expand_13", + "unsqueeze_46", + "transpose_95", + "matmul_11", + "mul_321", + "add_253", + "softmax_5", + "matmul_12" + ] + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_1" + ], + "source": "b_model_model_language_model_rotary_emb_inv_freq" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_2" + ], + "source": "b_model_model_language_model_rotary_emb_inv_freq" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_3" + ], + "source": "b_model_model_language_model_rotary_emb_inv_freq" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_4" + ], + "source": "slice_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as" + ], + "source": "mul_3" + }, + { + "pattern": "elided_view", + "nodes": [ + "view" + ], + "source": "linear_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze" + ], + "source": "p_model_model_language_model_layers_0_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_5" + ], + "source": "p_model_model_language_model_layers_0_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_1" + ], + "source": "slice_13" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_2" + ], + "source": "slice_14" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_3" + ], + "source": "slice_15" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone" + ], + "source": "transpose_15" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_4" + ], + "source": "transpose_15" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_5" + ], + "source": "linear_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_6" + ], + "source": "mul_11" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_1" + ], + "source": "mul_13" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_2" + ], + "source": "mul_16" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_7" + ], + "source": "linear_9" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_1" + ], + "source": "p_model_model_language_model_layers_1_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_6" + ], + "source": "p_model_model_language_model_layers_1_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_8" + ], + "source": "slice_21" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_9" + ], + "source": "slice_22" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_10" + ], + "source": "slice_23" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_1" + ], + "source": "transpose_18" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_11" + ], + "source": "transpose_18" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_12" + ], + "source": "linear_9" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_13" + ], + "source": "mul_24" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_3" + ], + "source": "mul_26" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_4" + ], + "source": "mul_29" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_14" + ], + "source": "linear_17" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_2" + ], + "source": "p_model_model_language_model_layers_2_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_7" + ], + "source": "p_model_model_language_model_layers_2_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_15" + ], + "source": "slice_29" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_16" + ], + "source": "slice_30" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_17" + ], + "source": "slice_31" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_2" + ], + "source": "transpose_21" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_18" + ], + "source": "transpose_21" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_19" + ], + "source": "linear_17" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_20" + ], + "source": "mul_37" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_5" + ], + "source": "mul_39" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_6" + ], + "source": "mul_42" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_21" + ], + "source": "linear_24" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_3" + ], + "source": "slice_35" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view" + ], + "source": "slice_35" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_22" + ], + "source": "slice_34" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_7" + ], + "source": "mul_44" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_23" + ], + "source": "linear_25" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_8" + ], + "source": "mul_46" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_24" + ], + "source": "linear_26" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_8" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_9" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_6" + ], + "source": "transpose_26" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_25" + ], + "source": "transpose_26" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_9" + ], + "source": "mul_54" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_10" + ], + "source": "mul_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_26" + ], + "source": "linear_32" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_3" + ], + "source": "p_model_model_language_model_layers_4_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_12" + ], + "source": "p_model_model_language_model_layers_4_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_27" + ], + "source": "slice_47" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_28" + ], + "source": "slice_48" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_29" + ], + "source": "slice_49" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_7" + ], + "source": "transpose_29" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_30" + ], + "source": "transpose_29" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_31" + ], + "source": "linear_32" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_32" + ], + "source": "mul_65" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_11" + ], + "source": "mul_67" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_12" + ], + "source": "mul_70" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_33" + ], + "source": "linear_40" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_4" + ], + "source": "p_model_model_language_model_layers_5_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_13" + ], + "source": "p_model_model_language_model_layers_5_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_34" + ], + "source": "slice_55" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_35" + ], + "source": "slice_56" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_36" + ], + "source": "slice_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_8" + ], + "source": "transpose_32" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_37" + ], + "source": "transpose_32" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_38" + ], + "source": "linear_40" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_39" + ], + "source": "mul_78" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_13" + ], + "source": "mul_80" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_14" + ], + "source": "mul_83" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_40" + ], + "source": "linear_48" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_5" + ], + "source": "p_model_model_language_model_layers_6_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_14" + ], + "source": "p_model_model_language_model_layers_6_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_41" + ], + "source": "slice_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_42" + ], + "source": "slice_64" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_43" + ], + "source": "slice_65" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_9" + ], + "source": "transpose_35" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_44" + ], + "source": "transpose_35" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_45" + ], + "source": "linear_48" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_46" + ], + "source": "mul_91" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_15" + ], + "source": "mul_93" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_16" + ], + "source": "mul_96" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_47" + ], + "source": "linear_55" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_10" + ], + "source": "slice_69" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_3" + ], + "source": "slice_69" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_48" + ], + "source": "slice_68" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_17" + ], + "source": "mul_98" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_49" + ], + "source": "linear_56" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_18" + ], + "source": "mul_100" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_50" + ], + "source": "linear_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_15" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_16" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_13" + ], + "source": "transpose_40" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_51" + ], + "source": "transpose_40" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_19" + ], + "source": "mul_108" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_20" + ], + "source": "mul_111" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_52" + ], + "source": "linear_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_6" + ], + "source": "p_model_model_language_model_layers_8_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_19" + ], + "source": "p_model_model_language_model_layers_8_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_53" + ], + "source": "slice_81" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_54" + ], + "source": "slice_82" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_55" + ], + "source": "slice_83" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_14" + ], + "source": "transpose_43" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_56" + ], + "source": "transpose_43" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_57" + ], + "source": "linear_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_58" + ], + "source": "mul_119" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_21" + ], + "source": "mul_121" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_22" + ], + "source": "mul_124" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_59" + ], + "source": "linear_71" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_7" + ], + "source": "p_model_model_language_model_layers_9_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_20" + ], + "source": "p_model_model_language_model_layers_9_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_60" + ], + "source": "slice_89" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_61" + ], + "source": "slice_90" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_62" + ], + "source": "slice_91" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_15" + ], + "source": "transpose_46" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_63" + ], + "source": "transpose_46" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_64" + ], + "source": "linear_71" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_65" + ], + "source": "mul_132" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_23" + ], + "source": "mul_134" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_24" + ], + "source": "mul_137" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_66" + ], + "source": "linear_79" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_8" + ], + "source": "p_model_model_language_model_layers_10_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_21" + ], + "source": "p_model_model_language_model_layers_10_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_67" + ], + "source": "slice_97" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_68" + ], + "source": "slice_98" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_69" + ], + "source": "slice_99" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_16" + ], + "source": "transpose_49" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_70" + ], + "source": "transpose_49" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_71" + ], + "source": "linear_79" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_72" + ], + "source": "mul_145" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_25" + ], + "source": "mul_147" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_26" + ], + "source": "mul_150" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_73" + ], + "source": "linear_86" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_17" + ], + "source": "slice_103" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_6" + ], + "source": "slice_103" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_74" + ], + "source": "slice_102" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_27" + ], + "source": "mul_152" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_75" + ], + "source": "linear_87" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_28" + ], + "source": "mul_154" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_76" + ], + "source": "linear_88" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_22" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_23" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_20" + ], + "source": "transpose_54" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_77" + ], + "source": "transpose_54" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_29" + ], + "source": "mul_162" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_30" + ], + "source": "mul_165" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_78" + ], + "source": "linear_94" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_9" + ], + "source": "p_model_model_language_model_layers_12_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_26" + ], + "source": "p_model_model_language_model_layers_12_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_79" + ], + "source": "slice_115" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_80" + ], + "source": "slice_116" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_81" + ], + "source": "slice_117" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_21" + ], + "source": "transpose_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_82" + ], + "source": "transpose_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_83" + ], + "source": "linear_94" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_84" + ], + "source": "mul_173" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_31" + ], + "source": "mul_175" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_32" + ], + "source": "mul_178" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_85" + ], + "source": "linear_102" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_10" + ], + "source": "p_model_model_language_model_layers_13_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_27" + ], + "source": "p_model_model_language_model_layers_13_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_86" + ], + "source": "slice_123" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_87" + ], + "source": "slice_124" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_88" + ], + "source": "slice_125" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_22" + ], + "source": "transpose_60" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_89" + ], + "source": "transpose_60" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_90" + ], + "source": "linear_102" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_91" + ], + "source": "mul_186" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_33" + ], + "source": "mul_188" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_34" + ], + "source": "mul_191" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_92" + ], + "source": "linear_110" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_11" + ], + "source": "p_model_model_language_model_layers_14_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_28" + ], + "source": "p_model_model_language_model_layers_14_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_93" + ], + "source": "slice_131" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_94" + ], + "source": "slice_132" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_95" + ], + "source": "slice_133" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_23" + ], + "source": "transpose_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_96" + ], + "source": "transpose_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_97" + ], + "source": "linear_110" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_98" + ], + "source": "mul_199" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_35" + ], + "source": "mul_201" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_36" + ], + "source": "mul_204" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_99" + ], + "source": "linear_117" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_24" + ], + "source": "slice_137" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_9" + ], + "source": "slice_137" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_100" + ], + "source": "slice_136" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_37" + ], + "source": "mul_206" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_101" + ], + "source": "linear_118" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_38" + ], + "source": "mul_208" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_102" + ], + "source": "linear_119" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_29" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_30" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_27" + ], + "source": "transpose_68" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_103" + ], + "source": "transpose_68" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_39" + ], + "source": "mul_216" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_40" + ], + "source": "mul_219" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_104" + ], + "source": "linear_125" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_12" + ], + "source": "p_model_model_language_model_layers_16_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_33" + ], + "source": "p_model_model_language_model_layers_16_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_105" + ], + "source": "slice_149" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_106" + ], + "source": "slice_150" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_107" + ], + "source": "slice_151" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_28" + ], + "source": "transpose_71" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_108" + ], + "source": "transpose_71" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_109" + ], + "source": "linear_125" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_110" + ], + "source": "mul_227" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_41" + ], + "source": "mul_229" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_42" + ], + "source": "mul_232" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_111" + ], + "source": "linear_133" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_13" + ], + "source": "p_model_model_language_model_layers_17_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_34" + ], + "source": "p_model_model_language_model_layers_17_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_112" + ], + "source": "slice_157" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_113" + ], + "source": "slice_158" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_114" + ], + "source": "slice_159" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_29" + ], + "source": "transpose_74" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_115" + ], + "source": "transpose_74" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_116" + ], + "source": "linear_133" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_117" + ], + "source": "mul_240" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_43" + ], + "source": "mul_242" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_44" + ], + "source": "mul_245" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_118" + ], + "source": "linear_141" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_14" + ], + "source": "p_model_model_language_model_layers_18_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_35" + ], + "source": "p_model_model_language_model_layers_18_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_119" + ], + "source": "slice_165" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_120" + ], + "source": "slice_166" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_121" + ], + "source": "slice_167" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_30" + ], + "source": "transpose_77" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_122" + ], + "source": "transpose_77" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_123" + ], + "source": "linear_141" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_124" + ], + "source": "mul_253" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_45" + ], + "source": "mul_255" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_46" + ], + "source": "mul_258" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_125" + ], + "source": "linear_148" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_31" + ], + "source": "slice_171" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_12" + ], + "source": "slice_171" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_126" + ], + "source": "slice_170" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_47" + ], + "source": "mul_260" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_127" + ], + "source": "linear_149" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_48" + ], + "source": "mul_262" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_128" + ], + "source": "linear_150" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_36" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_37" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_34" + ], + "source": "transpose_82" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_129" + ], + "source": "transpose_82" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_49" + ], + "source": "mul_270" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_50" + ], + "source": "mul_273" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_130" + ], + "source": "linear_156" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_15" + ], + "source": "p_model_model_language_model_layers_20_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_40" + ], + "source": "p_model_model_language_model_layers_20_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_131" + ], + "source": "slice_183" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_132" + ], + "source": "slice_184" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_133" + ], + "source": "slice_185" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_35" + ], + "source": "transpose_85" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_134" + ], + "source": "transpose_85" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_135" + ], + "source": "linear_156" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_136" + ], + "source": "mul_281" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_51" + ], + "source": "mul_283" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_52" + ], + "source": "mul_286" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_137" + ], + "source": "linear_164" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_16" + ], + "source": "p_model_model_language_model_layers_21_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_41" + ], + "source": "p_model_model_language_model_layers_21_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_138" + ], + "source": "slice_191" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_139" + ], + "source": "slice_192" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_140" + ], + "source": "slice_193" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_36" + ], + "source": "transpose_88" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_141" + ], + "source": "transpose_88" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_142" + ], + "source": "linear_164" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_143" + ], + "source": "mul_294" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_53" + ], + "source": "mul_296" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_54" + ], + "source": "mul_299" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_144" + ], + "source": "linear_172" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_17" + ], + "source": "p_model_model_language_model_layers_22_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_42" + ], + "source": "p_model_model_language_model_layers_22_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_145" + ], + "source": "slice_199" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_146" + ], + "source": "slice_200" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_147" + ], + "source": "slice_201" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_37" + ], + "source": "transpose_91" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_148" + ], + "source": "transpose_91" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_149" + ], + "source": "linear_172" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_150" + ], + "source": "mul_307" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_55" + ], + "source": "mul_309" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_56" + ], + "source": "mul_312" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_151" + ], + "source": "linear_179" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_38" + ], + "source": "slice_205" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_15" + ], + "source": "slice_205" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_152" + ], + "source": "slice_204" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_57" + ], + "source": "mul_314" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_153" + ], + "source": "linear_180" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_58" + ], + "source": "mul_316" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_154" + ], + "source": "linear_181" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_43" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_44" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_41" + ], + "source": "transpose_96" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_155" + ], + "source": "transpose_96" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_59" + ], + "source": "mul_324" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_60" + ], + "source": "mul_327" + }, + { + "pattern": "elided_view", + "nodes": [ + "alias" + ], + "source": "mul_327" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_10" + ], + "input": "states_0", + "output": "slice_10", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_16" + ], + "input": "states_1", + "output": "slice_16", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_18" + ], + "input": "states_2", + "output": "slice_18", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_24" + ], + "input": "states_3", + "output": "slice_24", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_26" + ], + "input": "states_4", + "output": "slice_26", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_32" + ], + "input": "states_5", + "output": "slice_32", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_44" + ], + "input": "states_8", + "output": "slice_44", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_50" + ], + "input": "states_9", + "output": "slice_50", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_52" + ], + "input": "states_10", + "output": "slice_52", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_58" + ], + "input": "states_11", + "output": "slice_58", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_60" + ], + "input": "states_12", + "output": "slice_60", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_66" + ], + "input": "states_13", + "output": "slice_66", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_78" + ], + "input": "states_16", + "output": "slice_78", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_84" + ], + "input": "states_17", + "output": "slice_84", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_86" + ], + "input": "states_18", + "output": "slice_86", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_92" + ], + "input": "states_19", + "output": "slice_92", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_94" + ], + "input": "states_20", + "output": "slice_94", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_100" + ], + "input": "states_21", + "output": "slice_100", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_112" + ], + "input": "states_24", + "output": "slice_112", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_118" + ], + "input": "states_25", + "output": "slice_118", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_120" + ], + "input": "states_26", + "output": "slice_120", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_126" + ], + "input": "states_27", + "output": "slice_126", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_128" + ], + "input": "states_28", + "output": "slice_128", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_134" + ], + "input": "states_29", + "output": "slice_134", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_146" + ], + "input": "states_32", + "output": "slice_146", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_152" + ], + "input": "states_33", + "output": "slice_152", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_154" + ], + "input": "states_34", + "output": "slice_154", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_160" + ], + "input": "states_35", + "output": "slice_160", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_162" + ], + "input": "states_36", + "output": "slice_162", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_168" + ], + "input": "states_37", + "output": "slice_168", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_180" + ], + "input": "states_40", + "output": "slice_180", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_186" + ], + "input": "states_41", + "output": "slice_186", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_188" + ], + "input": "states_42", + "output": "slice_188", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_194" + ], + "input": "states_43", + "output": "slice_194", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_196" + ], + "input": "states_44", + "output": "slice_196", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_202" + ], + "input": "states_45", + "output": "slice_202", + "kind": "recurrent" + } + ], + "partitions": [ + { + "node": "embedding", + "weight": "p_model_model_language_model_embed_tokens_weight", + "rowsPerShard": 131072, + "bindingLimitBytes": 134217728 + }, + { + "node": "linear_186", + "weight": "p_model_model_language_model_embed_tokens_weight", + "rowsPerShard": 131072, + "bindingLimitBytes": 134217728 + } + ] +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/entrypoints/prefill-64.json b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/entrypoints/prefill-64.json new file mode 100644 index 0000000000000000000000000000000000000000..6a000b461ad2bc0a5087f3e56348489c226d67e7 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/entrypoints/prefill-64.json @@ -0,0 +1,83074 @@ +{ + "inputs": [ + "input_ids", + "position_ids", + "attention_mask", + "states_0", + "states_1", + "states_2", + "states_3", + "states_4", + "states_5", + "states_6", + "states_7", + "states_8", + "states_9", + "states_10", + "states_11", + "states_12", + "states_13", + "states_14", + "states_15", + "states_16", + "states_17", + "states_18", + "states_19", + "states_20", + "states_21", + "states_22", + "states_23", + "states_24", + "states_25", + "states_26", + "states_27", + "states_28", + "states_29", + "states_30", + "states_31", + "states_32", + "states_33", + "states_34", + "states_35", + "states_36", + "states_37", + "states_38", + "states_39", + "states_40", + "states_41", + "states_42", + "states_43", + "states_44", + "states_45", + "states_46", + "states_47" + ], + "outputs": [ + "linear_186", + "slice_10", + "slice_16", + "slice_18", + "slice_24", + "slice_26", + "slice_32", + "cache_update", + "cache_update_1", + "slice_44", + "slice_50", + "slice_52", + "slice_58", + "slice_60", + "slice_66", + "cache_update_2", + "cache_update_3", + "slice_78", + "slice_84", + "slice_86", + "slice_92", + "slice_94", + "slice_100", + "cache_update_4", + "cache_update_5", + "slice_112", + "slice_118", + "slice_120", + "slice_126", + "slice_128", + "slice_134", + "cache_update_6", + "cache_update_7", + "slice_146", + "slice_152", + "slice_154", + "slice_160", + "slice_162", + "slice_168", + "cache_update_8", + "cache_update_9", + "slice_180", + "slice_186", + "slice_188", + "slice_194", + "slice_196", + "slice_202", + "cache_update_10", + "cache_update_11" + ], + "tensors": { + "p_model_model_language_model_embed_tokens_weight": { + "shape": [ + 248320, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 254279680, + "storageDtype": "uint32", + "parts": [ + { + "path": "weights/int8-422da3921400-000.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-001.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-002.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-003.bin", + "offset": 0, + "byteLength": 52953088 + } + ], + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_embed_tokens_weight__int8_scales", + "wordsPerRow": 256 + } + }, + "p_model_model_language_model_layers_0_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 38797312, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 38801408, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 38801472, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 38801536, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 45092992, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47190144, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47206528, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47222912, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47321216, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 47321728, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_0_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 49418880, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 49422976, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 51257984, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_0_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 53092992, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_1_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 54928000, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 54932096, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 54932160, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 54932224, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 61223680, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63320832, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63337216, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63353600, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63451904, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 63452416, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_1_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 65549568, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00007.bin", + "offset": 65553664, + "byteLength": 1555200 + }, + { + "path": "weights/00008.bin", + "offset": 0, + "byteLength": 279808 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 279808, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_1_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 2114816, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_2_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 3949824, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 3953920, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 3953984, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 3954048, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 10245504, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12342656, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12359040, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12375424, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12473728, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 12474240, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_2_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 14571392, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 14575488, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 16410496, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_2_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 18245504, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_3_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 20080512, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 20084608, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 24278912, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 24279936, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 24804224, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 24805248, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 25329536, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_3_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 27426688, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 27430784, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 29265792, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_3_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 31100800, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_4_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 32935808, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 32939904, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 32939968, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 32940032, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 39231488, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41328640, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41345024, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41361408, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41459712, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 41460224, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_4_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 43557376, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 43561472, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 45396480, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_4_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 47231488, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_5_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 49066496, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 49070592, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 49070656, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 49070720, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 55362176, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57459328, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57475712, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57492096, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57590400, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 57590912, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_5_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 59688064, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 59692160, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 61527168, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_5_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 63362176, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_6_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 65197184, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 65201280, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 65201344, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00008.bin", + "offset": 65201408, + "byteLength": 1907456 + }, + { + "path": "weights/00009.bin", + "offset": 0, + "byteLength": 4384000 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 4384000, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6481152, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6497536, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6513920, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6612224, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 6612736, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_6_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 8709888, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 8713984, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 10548992, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_6_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 12384000, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_7_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 14219008, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 14223104, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 18417408, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 18418432, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 18942720, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 18943744, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 19468032, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_7_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 21565184, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 21569280, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 23404288, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_7_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 25239296, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_8_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 27074304, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 27078400, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 27078464, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 27078528, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 33369984, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35467136, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35483520, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35499904, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35598208, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 35598720, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_8_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 37695872, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 37699968, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 39534976, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_8_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 41369984, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_9_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 43204992, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 43209088, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 43209152, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 43209216, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 49500672, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51597824, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51614208, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51630592, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51728896, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 51729408, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_9_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 53826560, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 53830656, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 55665664, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_9_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 57500672, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_10_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 59335680, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 59339776, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 59339840, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 59339904, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00009.bin", + "offset": 65631360, + "byteLength": 1477504 + }, + { + "path": "weights/00010.bin", + "offset": 0, + "byteLength": 619648 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 619648, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 636032, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 652416, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 750720, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 751232, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_10_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 2848384, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 2852480, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 4687488, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_10_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 6522496, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_11_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 8357504, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 8361600, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 12555904, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 12556928, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 13081216, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 13082240, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 13606528, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_11_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 15703680, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 15707776, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 17542784, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_11_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 19377792, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_12_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 21212800, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 21216896, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 21216960, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 21217024, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 27508480, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29605632, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29622016, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29638400, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29736704, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 29737216, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_12_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 31834368, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 31838464, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 33673472, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_12_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 35508480, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_13_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 37343488, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 37347584, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 37347648, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 37347712, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 43639168, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45736320, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45752704, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45769088, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45867392, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 45867904, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_13_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 47965056, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 47969152, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 49804160, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_13_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 51639168, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_14_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 53474176, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 53478272, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 53478336, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 53478400, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 59769856, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61867008, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61883392, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61899776, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61998080, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 61998592, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_14_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 64095744, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 64099840, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00010.bin", + "offset": 65934848, + "byteLength": 1174016 + }, + { + "path": "weights/00011.bin", + "offset": 0, + "byteLength": 660992 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_14_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 660992, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_15_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 2496000, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 2500096, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 6694400, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 6695424, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 7219712, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 7220736, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 7745024, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_15_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 9842176, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 9846272, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 11681280, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_15_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 13516288, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_16_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 15351296, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 15355392, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 15355456, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 15355520, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 21646976, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23744128, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23760512, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23776896, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23875200, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 23875712, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_16_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 25972864, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 25976960, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 27811968, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_16_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 29646976, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_17_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 31481984, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 31486080, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 31486144, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 31486208, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 37777664, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 39874816, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 39891200, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 39907584, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 40005888, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 40006400, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_17_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 42103552, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 42107648, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 43942656, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_17_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 45777664, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_18_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 47612672, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 47616768, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 47616832, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 47616896, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 53908352, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56005504, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56021888, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56038272, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56136576, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 56137088, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_18_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 58234240, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 58238336, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 60073344, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_18_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 61908352, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_19_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 63743360, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00011.bin", + "offset": 63747456, + "byteLength": 3361408 + }, + { + "path": "weights/00012.bin", + "offset": 0, + "byteLength": 832896 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 832896, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 833920, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 1358208, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 1359232, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 1883520, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_19_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 3980672, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 3984768, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 5819776, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_19_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 7654784, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_20_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 9489792, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 9493888, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 9493952, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 9494016, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 15785472, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 17882624, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 17899008, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 17915392, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 18013696, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 18014208, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_20_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 20111360, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 20115456, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 21950464, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_20_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 23785472, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_21_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 25620480, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 25624576, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 25624640, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 25624704, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 31916160, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34013312, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34029696, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34046080, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34144384, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 34144896, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_21_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 36242048, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 36246144, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 38081152, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_21_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 39916160, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_22_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 41751168, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_a_log": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 41755264, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_dt_bias": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 64, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 41755328, + "byteLength": 64 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight": { + "shape": [ + 6144, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 6291456, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 41755392, + "byteLength": 6291456 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight": { + "shape": [ + 2048, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 48046848, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50144000, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight": { + "shape": [ + 16, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16384, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50160384, + "byteLength": 16384 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_conv1d_weight": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 98304, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50176768, + "byteLength": 98304 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_norm_weight": { + "shape": [ + 128 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 512, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50275072, + "byteLength": 512 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_linear_attn_out_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 50275584, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_22_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 52372736, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 52376832, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 54211840, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_22_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 56046848, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_23_input_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 57881856, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_q_proj_weight": { + "shape": [ + 4096, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4194304, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_self_attn_q_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 57885952, + "byteLength": 4194304 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_q_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 62080256, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_k_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_self_attn_k_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 62081280, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_k_norm_weight": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1024, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 62605568, + "byteLength": 1024 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_v_proj_weight": { + "shape": [ + 512, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_self_attn_v_proj_weight__int8_scales", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 62606592, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_o_proj_weight": { + "shape": [ + 1024, + 2048 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2097152, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_self_attn_o_proj_weight__int8_scales", + "wordsPerRow": 512 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 63130880, + "byteLength": 2097152 + } + ] + }, + "p_model_model_language_model_layers_23_post_attention_layernorm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 65228032, + "byteLength": 4096 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_gate_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_mlp_gate_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 65232128, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_up_proj_weight": { + "shape": [ + 3584, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales", + "wordsPerRow": 128 + }, + "parts": [ + { + "path": "weights/00012.bin", + "offset": 67067136, + "byteLength": 41728 + }, + { + "path": "weights/00013.bin", + "offset": 0, + "byteLength": 1793280 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_down_proj_weight": { + "shape": [ + 1024, + 3584 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 1835008, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int4", + "groupSize": 32, + "scales": "p_model_model_language_model_layers_23_mlp_down_proj_weight__int4_scales", + "wordsPerRow": 448 + }, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 1793280, + "byteLength": 1835008 + } + ] + }, + "p_model_model_language_model_norm_weight": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 4096, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 3628288, + "byteLength": 4096 + } + ] + }, + "b_model_model_language_model_rotary_emb_inv_freq": { + "shape": [ + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 128, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 3632384, + "byteLength": 128 + } + ] + }, + "input_ids": { + "shape": [ + 1, + 64 + ], + "dtype": "int32", + "role": "input", + "byteLength": 256, + "indexBounds": [ + 0, + 248320 + ] + }, + "position_ids": { + "shape": [ + 1, + 64 + ], + "dtype": "int32", + "role": "input", + "byteLength": 256, + "indexBounds": [ + 0, + 4096 + ] + }, + "attention_mask": { + "shape": [ + 1, + 1, + 64, + 4096 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_0": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_1": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_2": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_3": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_4": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_5": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_6": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_7": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_8": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_9": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_10": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_11": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_12": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_13": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_14": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_15": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_16": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_17": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_18": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_19": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_20": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_21": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_22": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_23": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_24": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_25": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_26": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_27": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_28": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_29": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_30": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_31": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_32": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_33": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_34": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_35": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_36": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_37": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_38": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_39": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_40": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_41": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_42": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_43": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_44": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "input", + "byteLength": 98304 + }, + "states_45": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "input", + "byteLength": 1048576 + }, + "states_46": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "states_47": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "input", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "embedding": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "unsqueeze": { + "shape": [ + 1, + 1, + 64 + ], + "dtype": "int32", + "role": "value", + "byteLength": 256 + }, + "expand": { + "shape": [ + 4, + 1, + 64 + ], + "dtype": "int32", + "role": "value", + "byteLength": 1024 + }, + "slice_1": { + "shape": [ + 3, + 1, + 64 + ], + "dtype": "int32", + "role": "value", + "byteLength": 768 + }, + "unsqueeze_1": { + "shape": [ + 1, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "unsqueeze_2": { + "shape": [ + 1, + 1, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "unsqueeze_3": { + "shape": [ + 1, + 1, + 32, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 128 + }, + "expand_1": { + "shape": [ + 3, + 1, + 32, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 384 + }, + "unsqueeze_4": { + "shape": [ + 3, + 1, + 1, + 64 + ], + "dtype": "int32", + "role": "value", + "byteLength": 768 + }, + "_to_copy": { + "shape": [ + 3, + 1, + 1, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 768 + }, + "matmul": { + "shape": [ + 3, + 1, + 32, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "transpose": { + "shape": [ + 3, + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "select": { + "shape": [ + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "select_1": { + "shape": [ + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_2": { + "shape": [ + 1, + 64, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2816 + }, + "slice_3": { + "shape": [ + 1, + 64, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2816 + }, + "copy": { + "shape": [ + 1, + 64, + 11 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2816 + }, + "transpose_1": { + "shape": [ + 3, + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "select_2": { + "shape": [ + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_scatter": { + "shape": [ + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "select_scatter": { + "shape": [ + 3, + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "transpose_2": { + "shape": [ + 3, + 1, + 32, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "transpose_7": { + "shape": [ + 3, + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "select_7": { + "shape": [ + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_7": { + "shape": [ + 1, + 64, + 10 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2560 + }, + "transpose_8": { + "shape": [ + 3, + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "select_8": { + "shape": [ + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_8": { + "shape": [ + 1, + 64, + 10 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2560 + }, + "copy_1": { + "shape": [ + 1, + 64, + 10 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2560 + }, + "transpose_9": { + "shape": [ + 3, + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "select_9": { + "shape": [ + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "slice_scatter_1": { + "shape": [ + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "select_scatter_1": { + "shape": [ + 3, + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "transpose_10": { + "shape": [ + 3, + 1, + 32, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "transpose_12": { + "shape": [ + 3, + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 24576 + }, + "select_11": { + "shape": [ + 1, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8192 + }, + "cat": { + "shape": [ + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "cos": { + "shape": [ + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mul": { + "shape": [ + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "sin": { + "shape": [ + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "mul_1": { + "shape": [ + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "pow_1": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_2": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_1": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_3": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_13": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_1": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_2": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_3": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_1": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_10": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_5": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_11": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_12": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_14": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_13": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_14": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_15": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_1": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_2": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_3": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_2": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_4": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_5": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_1": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_3": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_1": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_6": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_7": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_2": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_4": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_2": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_8": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_16": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_17": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_15": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_4": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_5": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_2": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_1": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_5": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_3": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_9": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_10": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_1": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_11": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_6": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_4": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_6": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_3": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_2": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_7": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_4": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_12": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_8": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_13": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_1": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_5": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_2": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_6": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_14": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_7": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_9": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_4": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_3": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_10": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_5": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_15": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_11": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_16": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_2": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_8": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_16": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_9": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_7": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_10": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_11": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_2": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_18": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_1": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_6": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_1": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_19": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_3": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_20": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_17": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_21": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_22": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_23": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_8": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_9": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_10": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_1": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_1": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_1": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_12": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_1": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_17": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_18": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_3": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_13": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_6": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_19": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_20": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_4": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_14": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_7": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_21": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_1": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_24": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_25": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_18": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_1": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_11": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_12": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_5": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_4": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_15": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_8": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_22": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_23": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_4": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_24": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_13": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_12": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_16": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_6": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_5": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_17": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_9": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_25": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_18": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_26": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_3": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_13": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_5": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_14": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_27": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_15": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_19": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_7": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_6": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_20": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_10": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_28": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_21": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_29": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_4": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_16": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_19": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_17": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_14": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_18": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_19": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_3": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_26": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_2": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_7": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_2": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_27": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_6": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_28": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_20": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_29": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_30": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_31": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_15": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_16": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_17": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_2": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_2": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_2": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_22": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_2": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_30": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_31": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_5": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_23": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_11": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_32": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_33": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_6": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_24": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_12": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_34": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_2": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_32": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_33": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_21": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_2": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_18": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_19": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_8": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_7": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_25": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_13": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_35": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_36": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_7": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_37": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_20": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_20": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_26": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_9": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_8": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_27": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_14": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_38": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_28": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_39": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_5": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_21": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_8": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_22": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_40": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_23": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_29": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_10": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_9": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_30": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_15": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_41": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_31": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_42": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_6": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_24": { + "shape": [ + 1, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "view_21": { + "shape": [ + 1, + 64, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_34": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_35": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_3": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "_unsafe_view": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_22": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_11": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_10": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "add_32": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "rsqrt_16": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "mul_43": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "add_33": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_44": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "type_as_7": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_22": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_25": { + "shape": [ + 1, + 64, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_23": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_12": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_11": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "add_34": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "rsqrt_17": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "mul_45": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_35": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_46": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "type_as_8": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_23": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_26": { + "shape": [ + 1, + 64, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_24": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_24": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "unsqueeze_8": { + "shape": [ + 1, + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "unsqueeze_9": { + "shape": [ + 1, + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_36": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_37": { + "shape": [ + 1, + 8, + 64, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_38": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_39": { + "shape": [ + 1, + 2, + 64, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "mul_47": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_40": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "slice_41": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "neg_3": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "cat_4": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_48": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_36": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_49": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_42": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_43": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "neg_4": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "cat_5": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_50": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_37": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cat_6": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "cat_7": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "cache_update": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_1": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_10": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_2": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_4": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_1": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_11": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_3": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_5": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_2": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_25": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_1": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "mul_51": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "add_38": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "softmax": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "matmul_2": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_26": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_6": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_25": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_3": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_52": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_27": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_39": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_13": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_12": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_40": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_18": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_53": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_41": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_54": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_9": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_28": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_9": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_29": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_55": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_30": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_42": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_14": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_13": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_43": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_19": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_56": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_44": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_57": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_10": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_31": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_27": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_32": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_26": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_33": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_34": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_8": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_44": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_3": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_12": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_3": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_45": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_10": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_46": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_28": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_47": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_48": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_49": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_27": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_28": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_29": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_4": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_3": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_5": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_45": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_3": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_58": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_59": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_7": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_46": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_20": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_60": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_61": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_8": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_47": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_21": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_62": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_3": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_50": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_51": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_29": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_7": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_30": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_31": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_15": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_14": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_48": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_22": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_63": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_64": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_11": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_65": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_32": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_35": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_49": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_16": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_15": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_50": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_23": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_66": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_51": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_67": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_11": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_36": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_12": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_37": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_68": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_38": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_52": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_17": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_16": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_53": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_24": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_69": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_54": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_70": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_12": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_39": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_30": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_40": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_33": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_41": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_42": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_9": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_52": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_4": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_13": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_4": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_53": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_13": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_54": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_31": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_55": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_56": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_57": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_34": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_35": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_36": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_5": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_4": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_6": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_55": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_4": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_71": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_72": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_9": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_56": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_25": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_73": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_74": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_10": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_57": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_26": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_75": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_4": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_58": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_59": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_32": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_8": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_37": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_38": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_18": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_17": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_58": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_27": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_76": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_77": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_14": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_78": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_39": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_43": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_59": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_19": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_18": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_60": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_28": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_79": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_61": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_80": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_13": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_44": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_15": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_45": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_81": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_46": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_62": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_20": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_19": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_63": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_29": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_82": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_64": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_83": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_14": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_47": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_33": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_48": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_40": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_49": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_50": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_10": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_60": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_5": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_14": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_5": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_61": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_16": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_62": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_34": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_63": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_64": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_65": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_41": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_42": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_43": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_6": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_5": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_7": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_65": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_5": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_84": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_85": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_11": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_66": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_30": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_86": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_87": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_12": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_67": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_31": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_88": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_5": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_66": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_67": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_35": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_9": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_44": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_45": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_21": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_20": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_68": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_32": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_89": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_90": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_17": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_91": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_46": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_51": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_69": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_22": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_21": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_70": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_33": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_92": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_71": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_93": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_15": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_52": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_18": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_53": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_94": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_54": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_72": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_23": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_22": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_73": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_34": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_95": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_74": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_96": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_16": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_55": { + "shape": [ + 1, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "view_47": { + "shape": [ + 1, + 64, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_68": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_69": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_10": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "_unsafe_view_3": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_48": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_24": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_23": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "add_75": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "rsqrt_35": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "mul_97": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "add_76": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_98": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "type_as_17": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_36": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_56": { + "shape": [ + 1, + 64, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_49": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_25": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_24": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "add_77": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "rsqrt_36": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "mul_99": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_78": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_100": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "type_as_18": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_37": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_57": { + "shape": [ + 1, + 64, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_50": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_38": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "unsqueeze_15": { + "shape": [ + 1, + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "unsqueeze_16": { + "shape": [ + 1, + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_70": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_71": { + "shape": [ + 1, + 8, + 64, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_72": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_73": { + "shape": [ + 1, + 2, + 64, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "mul_101": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_74": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "slice_75": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "neg_8": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "cat_11": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_102": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_79": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_103": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_76": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_77": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "neg_9": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "cat_12": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_104": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_80": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cat_13": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "cat_14": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "cache_update_2": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_3": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_17": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_4": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_11": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_4": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_18": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_5": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_12": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_5": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_39": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_3": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "mul_105": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "add_81": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "softmax_1": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "matmul_4": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_40": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_13": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_51": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_7": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_106": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_58": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_82": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_26": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_25": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_83": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_37": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_107": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_84": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_108": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_19": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_59": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_19": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_60": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_109": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_61": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_85": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_27": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_26": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_86": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_38": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_110": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_87": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_111": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_20": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_62": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_41": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_63": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_52": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_64": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_65": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_15": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_78": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_6": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_19": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_6": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_79": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_20": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_80": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_42": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_81": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_82": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_83": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_53": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_54": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_55": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_8": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_6": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_10": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_88": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_6": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_112": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_113": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_13": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_89": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_39": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_114": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_115": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_14": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_90": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_40": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_116": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_6": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_84": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_85": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_43": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_14": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_56": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_57": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_28": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_27": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_91": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_41": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_117": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_118": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_21": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_119": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_58": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_66": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_92": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_29": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_28": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_93": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_42": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_120": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_94": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_121": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_21": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_67": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_22": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_68": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_122": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_69": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_95": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_30": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_29": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_96": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_43": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_123": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_97": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_124": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_22": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_70": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_44": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_71": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_59": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_72": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_73": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_16": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_86": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_7": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_20": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_7": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_87": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_23": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_88": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_45": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_89": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_90": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_91": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_60": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_61": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_62": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_9": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_7": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_11": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_98": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_7": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_125": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_126": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_15": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_99": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_44": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_127": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_128": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_16": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_100": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_45": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_129": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_7": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_92": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_93": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_46": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_15": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_63": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_64": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_31": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_30": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_101": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_46": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_130": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_131": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_24": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_132": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_65": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_74": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_102": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_32": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_31": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_103": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_47": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_133": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_104": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_134": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_23": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_75": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_25": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_76": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_135": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_77": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_105": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_33": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_32": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_106": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_48": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_136": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_107": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_137": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_24": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_78": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_47": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_79": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_66": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_80": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_81": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_17": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_94": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_8": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_21": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_8": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_95": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_26": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_96": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_48": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_97": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_98": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_99": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_67": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_68": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_69": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_10": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_8": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_12": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_108": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_8": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_138": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_139": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_17": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_109": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_49": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_140": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_141": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_18": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_110": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_50": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_142": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_8": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_100": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_101": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_49": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_16": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_70": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_71": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_34": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_33": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_111": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_51": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_143": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_144": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_27": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_145": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_72": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_82": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_112": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_35": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_34": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_113": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_52": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_146": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_114": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_147": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_25": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_83": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_28": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_84": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_148": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_85": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_115": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_36": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_35": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_116": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_53": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_149": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_117": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_150": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_26": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_86": { + "shape": [ + 1, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "view_73": { + "shape": [ + 1, + 64, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_102": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_103": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_17": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "_unsafe_view_6": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_74": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_37": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_36": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "add_118": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "rsqrt_54": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "mul_151": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "add_119": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_152": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "type_as_27": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_50": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_87": { + "shape": [ + 1, + 64, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_75": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_38": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_37": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "add_120": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "rsqrt_55": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "mul_153": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_121": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_154": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "type_as_28": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_51": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_88": { + "shape": [ + 1, + 64, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_76": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_52": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "unsqueeze_22": { + "shape": [ + 1, + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "unsqueeze_23": { + "shape": [ + 1, + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_104": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_105": { + "shape": [ + 1, + 8, + 64, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_106": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_107": { + "shape": [ + 1, + 2, + 64, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "mul_155": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_108": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "slice_109": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "neg_13": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "cat_18": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_156": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_122": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_157": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_110": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_111": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "neg_14": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "cat_19": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_158": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_123": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cat_20": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "cat_21": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "cache_update_4": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_5": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_24": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_6": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_18": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_7": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_25": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_7": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_19": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_8": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_53": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_5": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "mul_159": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "add_124": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "softmax_2": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "matmul_6": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_54": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_20": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_77": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_11": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_160": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_89": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_125": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_39": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_38": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_126": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_56": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_161": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_127": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_162": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_29": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_90": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_29": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_91": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_163": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_92": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_128": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_40": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_39": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_129": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_57": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_164": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_130": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_165": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_30": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_93": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_55": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_94": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_78": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_95": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_96": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_22": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_112": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_9": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_26": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_9": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_113": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_30": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_114": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_56": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_115": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_116": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_117": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_79": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_80": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_81": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_12": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_9": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_15": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_131": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_9": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_166": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_167": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_19": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_132": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_58": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_168": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_169": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_20": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_133": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_59": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_170": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_9": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_118": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_119": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_57": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_21": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_82": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_83": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_41": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_40": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_134": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_60": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_171": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_172": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_31": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_173": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_84": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_97": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_135": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_42": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_41": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_136": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_61": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_174": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_137": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_175": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_31": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_98": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_32": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_99": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_176": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_100": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_138": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_43": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_42": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_139": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_62": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_177": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_140": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_178": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_32": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_101": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_58": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_102": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_85": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_103": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_104": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_23": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_120": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_10": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_27": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_10": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_121": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_33": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_122": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_59": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_123": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_124": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_125": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_86": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_87": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_88": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_13": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_10": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_16": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_141": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_10": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_179": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_180": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_21": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_142": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_63": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_181": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_182": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_22": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_143": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_64": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_183": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_10": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_126": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_127": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_60": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_22": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_89": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_90": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_44": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_43": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_144": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_65": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_184": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_185": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_34": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_186": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_91": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_105": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_145": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_45": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_44": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_146": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_66": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_187": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_147": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_188": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_33": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_106": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_35": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_107": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_189": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_108": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_148": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_46": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_45": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_149": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_67": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_190": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_150": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_191": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_34": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_109": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_61": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_110": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_92": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_111": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_112": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_24": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_128": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_11": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_28": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_11": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_129": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_36": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_130": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_62": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_131": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_132": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_133": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_93": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_94": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_95": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_14": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_11": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_17": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_151": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_11": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_192": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_193": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_23": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_152": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_68": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_194": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_195": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_24": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_153": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_69": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_196": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_11": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_134": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_135": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_63": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_23": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_96": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_97": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_47": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_46": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_154": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_70": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_197": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_198": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_37": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_199": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_98": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_113": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_155": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_48": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_47": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_156": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_71": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_200": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_157": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_201": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_35": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_114": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_38": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_115": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_202": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_116": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_158": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_49": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_48": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_159": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_72": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_203": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_160": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_204": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_36": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_117": { + "shape": [ + 1, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "view_99": { + "shape": [ + 1, + 64, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_136": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_137": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_24": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "_unsafe_view_9": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_100": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_50": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_49": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "add_161": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "rsqrt_73": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "mul_205": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "add_162": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_206": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "type_as_37": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_64": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_118": { + "shape": [ + 1, + 64, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_101": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_51": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_50": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "add_163": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "rsqrt_74": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "mul_207": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_164": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_208": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "type_as_38": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_65": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_119": { + "shape": [ + 1, + 64, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_102": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_66": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "unsqueeze_29": { + "shape": [ + 1, + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "unsqueeze_30": { + "shape": [ + 1, + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_138": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_139": { + "shape": [ + 1, + 8, + 64, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_140": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_141": { + "shape": [ + 1, + 2, + 64, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "mul_209": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_142": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "slice_143": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "neg_18": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "cat_25": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_210": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_165": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_211": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_144": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_145": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "neg_19": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "cat_26": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_212": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_166": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cat_27": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "cat_28": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "cache_update_6": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_7": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_31": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_8": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_25": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_10": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_32": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_9": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_26": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_11": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_67": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_7": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "mul_213": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "add_167": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "softmax_3": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "matmul_8": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_68": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_27": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_103": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_15": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_214": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_120": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_168": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_52": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_51": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_169": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_75": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_215": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_170": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_216": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_39": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_121": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_39": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_122": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_217": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_123": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_171": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_53": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_52": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_172": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_76": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_218": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_173": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_219": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_40": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_124": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_69": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_125": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_104": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_126": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_127": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_29": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_146": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_12": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_33": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_12": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_147": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_40": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_148": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_70": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_149": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_150": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_151": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_105": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_106": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_107": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_16": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_12": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_20": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_174": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_12": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_220": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_221": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_25": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_175": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_77": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_222": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_223": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_26": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_176": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_78": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_224": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_12": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_152": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_153": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_71": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_28": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_108": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_109": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_54": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_53": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_177": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_79": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_225": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_226": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_41": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_227": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_110": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_128": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_178": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_55": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_54": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_179": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_80": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_228": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_180": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_229": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_41": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_129": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_42": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_130": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_230": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_131": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_181": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_56": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_55": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_182": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_81": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_231": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_183": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_232": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_42": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_132": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_72": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_133": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_111": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_134": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_135": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_30": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_154": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_13": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_34": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_13": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_155": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_43": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_156": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_73": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_157": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_158": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_159": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_112": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_113": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_114": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_17": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_13": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_21": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_184": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_13": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_233": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_234": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_27": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_185": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_82": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_235": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_236": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_28": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_186": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_83": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_237": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_13": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_160": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_161": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_74": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_29": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_115": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_116": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_57": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_56": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_187": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_84": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_238": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_239": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_44": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_240": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_117": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_136": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_188": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_58": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_57": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_189": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_85": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_241": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_190": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_242": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_43": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_137": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_45": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_138": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_243": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_139": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_191": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_59": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_58": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_192": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_86": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_244": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_193": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_245": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_44": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_140": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_75": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_141": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_118": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_142": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_143": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_31": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_162": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_14": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_35": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_14": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_163": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_46": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_164": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_76": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_165": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_166": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_167": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_119": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_120": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_121": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_18": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_14": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_22": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_194": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_14": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_246": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_247": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_29": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_195": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_87": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_248": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_249": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_30": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_196": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_88": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_250": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_14": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_168": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_169": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_77": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_30": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_122": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_123": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_60": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_59": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_197": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_89": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_251": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_252": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_47": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_253": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_124": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_144": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_198": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_61": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_60": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_199": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_90": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_254": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_200": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_255": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_45": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_145": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_48": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_146": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_256": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_147": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_201": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_62": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_61": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_202": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_91": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_257": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_203": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_258": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_46": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_148": { + "shape": [ + 1, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "view_125": { + "shape": [ + 1, + 64, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_170": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_171": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_31": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "_unsafe_view_12": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_126": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_63": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_62": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "add_204": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "rsqrt_92": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "mul_259": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "add_205": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_260": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "type_as_47": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_78": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_149": { + "shape": [ + 1, + 64, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_127": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_64": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_63": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "add_206": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "rsqrt_93": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "mul_261": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_207": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_262": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "type_as_48": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_79": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_150": { + "shape": [ + 1, + 64, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_128": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_80": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "unsqueeze_36": { + "shape": [ + 1, + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "unsqueeze_37": { + "shape": [ + 1, + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_172": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_173": { + "shape": [ + 1, + 8, + 64, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_174": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_175": { + "shape": [ + 1, + 2, + 64, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "mul_263": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_176": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "slice_177": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "neg_23": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "cat_32": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_264": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_208": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_265": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_178": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_179": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "neg_24": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "cat_33": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_266": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_209": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cat_34": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "cat_35": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "cache_update_8": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_9": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_38": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_10": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_32": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_13": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_39": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_11": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_33": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_14": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_81": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_9": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "mul_267": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "add_210": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "softmax_4": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "matmul_10": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_82": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_34": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_129": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_19": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_268": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_151": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_211": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_65": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_64": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_212": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_94": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_269": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_213": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_270": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_49": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_152": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_49": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_153": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_271": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_154": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_214": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_66": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_65": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_215": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_95": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_272": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_216": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_273": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_50": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_155": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_83": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_156": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_130": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_157": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_158": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_36": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_180": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_15": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_40": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_15": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_181": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_50": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_182": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_84": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_183": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_184": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_185": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_131": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_132": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_133": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_20": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_15": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_25": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_217": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_15": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_274": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_275": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_31": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_218": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_96": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_276": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_277": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_32": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_219": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_97": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_278": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_15": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_186": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_187": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_85": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_35": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_134": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_135": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_67": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_66": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_220": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_98": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_279": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_280": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_51": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_281": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_136": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_159": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_221": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_68": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_67": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_222": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_99": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_282": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_223": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_283": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_51": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_160": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_52": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_161": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_284": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_162": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_224": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_69": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_68": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_225": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_100": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_285": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_226": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_286": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_52": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_163": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_86": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_164": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_137": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_165": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_166": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_37": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_188": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_16": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_41": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_16": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_189": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_53": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_190": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_87": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_191": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_192": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_193": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_138": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_139": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_140": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_21": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_16": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_26": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_227": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_16": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_287": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_288": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_33": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_228": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_101": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_289": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_290": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_34": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_229": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_102": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_291": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_16": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_194": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_195": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_88": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_36": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_141": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_142": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_70": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_69": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_230": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_103": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_292": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_293": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_54": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_294": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_143": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_167": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_231": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_71": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_70": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_232": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_104": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_295": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_233": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_296": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_53": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_168": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_55": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_169": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_297": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_170": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_234": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_72": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_71": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_235": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_105": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_298": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_236": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_299": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_54": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_171": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_89": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "linear_172": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_144": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_173": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "linear_174": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "cat_38": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_196": { + "shape": [ + 1, + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "squeeze_17": { + "shape": [ + 6144, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "unsqueeze_42": { + "shape": [ + 6144, + 1, + 4 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "conv1d_17": { + "shape": [ + 1, + 6144, + 71 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1744896 + }, + "slice_197": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "silu_56": { + "shape": [ + 1, + 6144, + 68 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1671168 + }, + "slice_198": { + "shape": [ + 1, + 6144, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "transpose_90": { + "shape": [ + 1, + 64, + 6144 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_199": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_200": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_201": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_145": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_146": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_147": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_22": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "exp_17": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "neg_27": { + "shape": [ + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 64 + }, + "add_237": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "softplus_17": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_300": { + "shape": [ + 1, + 64, + 16 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_301": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_35": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_238": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_106": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_302": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_303": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sum_36": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_239": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_107": { + "shape": [ + 1, + 64, + 16, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_304": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "normalized_gated_delta_rule_17": { + "shape": [ + 1, + 16, + 192, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1572864 + }, + "slice_202": { + "shape": [ + 1, + 16, + 128, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_203": { + "shape": [ + 1, + 16, + 64, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_91": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_37": { + "shape": [ + 1, + 64, + 16, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_148": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_149": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_73": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_72": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "add_240": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "rsqrt_108": { + "shape": [ + 1024, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_305": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_306": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "silu_57": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_307": { + "shape": [ + 1024, + 128 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_150": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_175": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_241": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_74": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_73": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_242": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_109": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_308": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_243": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_309": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_55": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_176": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_58": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_177": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_310": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_178": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_244": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_75": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_74": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_245": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_110": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_311": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_246": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_312": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_56": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_179": { + "shape": [ + 1, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "view_151": { + "shape": [ + 1, + 64, + 8, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1048576 + }, + "slice_204": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "slice_205": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_38": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "_unsafe_view_15": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_152": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "pow_76": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mean_75": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "add_247": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "rsqrt_111": { + "shape": [ + 1, + 64, + 8, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 2048 + }, + "mul_313": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "add_248": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_314": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "type_as_57": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_92": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_180": { + "shape": [ + 1, + 64, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_153": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "pow_77": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mean_76": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "add_249": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "rsqrt_112": { + "shape": [ + 1, + 64, + 2, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 512 + }, + "mul_315": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_250": { + "shape": [ + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 1024 + }, + "mul_316": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "type_as_58": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_93": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "linear_181": { + "shape": [ + 1, + 64, + 512 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "view_154": { + "shape": [ + 1, + 64, + 2, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "transpose_94": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "unsqueeze_43": { + "shape": [ + 1, + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "unsqueeze_44": { + "shape": [ + 1, + 1, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_206": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_207": { + "shape": [ + 1, + 8, + 64, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 393216 + }, + "slice_208": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_209": { + "shape": [ + 1, + 2, + 64, + 192 + ], + "dtype": "float32", + "role": "value", + "byteLength": 98304 + }, + "mul_317": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "slice_210": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "slice_211": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "neg_28": { + "shape": [ + 1, + 8, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 65536 + }, + "cat_39": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_318": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "add_251": { + "shape": [ + 1, + 8, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "mul_319": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "slice_212": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "slice_213": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "neg_29": { + "shape": [ + 1, + 2, + 64, + 32 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16384 + }, + "cat_40": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "mul_320": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "add_252": { + "shape": [ + 1, + 2, + 64, + 64 + ], + "dtype": "float32", + "role": "value", + "byteLength": 32768 + }, + "cat_41": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "cat_42": { + "shape": [ + 1, + 2, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 131072 + }, + "cache_update_10": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "cache_update_11": { + "shape": [ + 1, + 2, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608, + "storageOrder": [ + 2, + 0, + 1, + 3 + ] + }, + "unsqueeze_45": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_12": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_39": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_16": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "unsqueeze_46": { + "shape": [ + 1, + 2, + 1, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "expand_13": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "clone_40": { + "shape": [ + 1, + 2, + 4, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "_unsafe_view_17": { + "shape": [ + 1, + 8, + 4096, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "transpose_95": { + "shape": [ + 1, + 8, + 256, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "matmul_11": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "mul_321": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "add_253": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "softmax_5": { + "shape": [ + 1, + 8, + 64, + 4096 + ], + "dtype": "float32", + "role": "value", + "byteLength": 8388608 + }, + "matmul_12": { + "shape": [ + 1, + 8, + 64, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "transpose_96": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "clone_41": { + "shape": [ + 1, + 64, + 8, + 256 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "view_155": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "sigmoid_23": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "mul_322": { + "shape": [ + 1, + 64, + 2048 + ], + "dtype": "float32", + "role": "value", + "byteLength": 524288 + }, + "linear_182": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_254": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_78": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_77": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_255": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_113": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_323": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_256": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_324": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_59": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_183": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "silu_59": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_184": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "mul_325": { + "shape": [ + 1, + 64, + 3584 + ], + "dtype": "float32", + "role": "value", + "byteLength": 917504 + }, + "linear_185": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_257": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "pow_79": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "mean_78": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "add_258": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "rsqrt_114": { + "shape": [ + 1, + 64, + 1 + ], + "dtype": "float32", + "role": "value", + "byteLength": 256 + }, + "mul_326": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "add_259": { + "shape": [ + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 4096 + }, + "mul_327": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "type_as_60": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "alias": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_186": { + "shape": [ + 1, + 64, + 248320 + ], + "dtype": "float32", + "role": "value", + "byteLength": 63569920 + }, + "p_model_model_language_model_layers_0_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 3632512, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 4091264, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 4550016, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 5008768, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 5467520, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 5926272, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 6385024, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 6843776, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 7302528, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 7761280, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 8220032, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 8678784, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 9137536, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 9596288, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 10055040, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 10513792, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 10972544, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 11431296, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 11890048, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 12348800, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 12807552, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 13266304, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 13725056, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 14183808, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 14642560, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 15101312, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 15560064, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 16018816, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 16477568, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 16936320, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 17395072, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 17853824, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 18312576, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 18771328, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 19230080, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 19688832, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 20147584, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 20606336, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 21065088, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 21523840, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 21982592, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 22441344, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 22900096, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 23358848, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 23817600, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 24276352, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 24735104, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 25193856, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 25652608, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 26111360, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 26570112, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 27028864, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 27487616, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 27946368, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 28405120, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 28863872, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 29322624, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 29781376, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 30240128, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 30698880, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 31157632, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 31616384, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 32075136, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 32533888, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 32992640, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 33451392, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 33910144, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 34368896, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 34827648, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_down_proj_weight__int4_scales": { + "shape": [ + 1024, + 112 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 35286400, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_gate_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 35745152, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales": { + "shape": [ + 3584, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 458752, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 36203904, + "byteLength": 458752 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 36662656, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 36664704, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 36666752, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37453184, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37715328, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37977472, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37979520, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 37981568, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 38768000, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 39030144, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 39292288, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 39357824, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 39619968, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_11_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 40144256, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 40209792, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 40211840, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 40213888, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41000320, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41262464, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41524608, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41526656, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 41528704, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42315136, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42577280, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42839424, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42841472, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 42843520, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 43629952, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 43892096, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 44154240, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 44219776, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 44481920, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_15_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45006208, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45071744, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45073792, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45075840, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 45862272, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 46124416, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 46386560, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 46388608, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 46390656, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47177088, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47439232, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47701376, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47703424, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 47705472, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 48491904, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 48754048, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49016192, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49081728, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49343872, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_19_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49868160, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49933696, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49935744, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 49937792, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 50724224, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 50986368, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 51248512, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 51250560, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 51252608, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52039040, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52301184, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52563328, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52565376, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 52567424, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53353856, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53616000, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53878144, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53880192, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 53882240, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 54668672, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 54930816, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 55192960, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 55258496, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 55520640, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_23_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56044928, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56110464, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56112512, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56114560, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 56900992, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 57163136, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 57425280, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 57490816, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 57752960, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_3_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 58277248, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 58342784, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 58344832, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 58346880, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59133312, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59395456, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59657600, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59659648, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 59661696, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60448128, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60710272, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60972416, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60974464, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 60976512, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 61762944, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 62025088, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_k_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 62287232, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_o_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 62352768, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_q_proj_weight__int8_scales": { + "shape": [ + 4096, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 524288, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 62614912, + "byteLength": 524288 + } + ] + }, + "p_model_model_language_model_layers_7_self_attn_v_proj_weight__int8_scales": { + "shape": [ + 512, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 65536, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63139200, + "byteLength": 65536 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63204736, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63206784, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63208832, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 63995264, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 64257408, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 64519552, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight__int8_scales": { + "shape": [ + 16, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 2048, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 64521600, + "byteLength": 2048 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight__int8_scales": { + "shape": [ + 6144, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 786432, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 64523648, + "byteLength": 786432 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight__int8_scales": { + "shape": [ + 2048, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 65310080, + "byteLength": 262144 + } + ] + }, + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight__int8_scales": { + "shape": [ + 1024, + 64 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 262144, + "parts": [ + { + "path": "weights/00013.bin", + "offset": 65572224, + "byteLength": 262144 + } + ] + }, + "embedding__part_0": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "embedding__part_1": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "embedding__part_2": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "embedding__part_3": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "embedding__sum_0": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "embedding__sum_1": { + "shape": [ + 1, + 64, + 1024 + ], + "dtype": "float32", + "role": "value", + "byteLength": 262144 + }, + "linear_186__part_0": { + "shape": [ + 1, + 64, + 131072 + ], + "dtype": "float32", + "role": "value", + "byteLength": 33554432 + }, + "linear_186__part_1": { + "shape": [ + 1, + 64, + 117248 + ], + "dtype": "float32", + "role": "value", + "byteLength": 30015488 + }, + "linear_186__part_2": { + "shape": [ + 1, + 64, + 65536 + ], + "dtype": "float32", + "role": "value", + "byteLength": 16777216 + }, + "linear_186__part_3": { + "shape": [ + 1, + 64, + 51712 + ], + "dtype": "float32", + "role": "value", + "byteLength": 13238272 + }, + "p_model_model_language_model_embed_tokens_weight__int8_scales": { + "shape": [ + 248320, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 31784960, + "parts": [ + { + "path": "weights/int8-422da3921400-scales-000.bin", + "offset": 0, + "byteLength": 31784960 + } + ] + }, + "p_model_model_language_model_embed_tokens_weight__rows_0_131072": { + "shape": [ + 131072, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 134217728, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 0 + }, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/int8-422da3921400-000.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-001.bin", + "offset": 0, + "byteLength": 67108864 + } + ] + }, + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072": { + "shape": [ + 131072, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 16777216, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight__int8_scales", + "offset": 0 + }, + "parts": [ + { + "path": "weights/int8-422da3921400-scales-000.bin", + "offset": 0, + "byteLength": 16777216 + } + ] + }, + "p_model_model_language_model_embed_tokens_weight__rows_131072_117248": { + "shape": [ + 117248, + 1024 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 120061952, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "offset": 134217728 + }, + "storageDtype": "uint32", + "quantization": { + "scheme": "symmetric-int8", + "groupSize": 32, + "scales": "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248", + "wordsPerRow": 256 + }, + "parts": [ + { + "path": "weights/int8-422da3921400-002.bin", + "offset": 0, + "byteLength": 67108864 + }, + { + "path": "weights/int8-422da3921400-003.bin", + "offset": 0, + "byteLength": 52953088 + } + ] + }, + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248": { + "shape": [ + 117248, + 32 + ], + "dtype": "float32", + "role": "constant", + "byteLength": 15007744, + "viewOf": { + "tensor": "p_model_model_language_model_embed_tokens_weight__int8_scales", + "offset": 16777216 + }, + "parts": [ + { + "path": "weights/int8-422da3921400-scales-000.bin", + "offset": 16777216, + "byteLength": 15007744 + } + ] + } + }, + "nodes": [ + { + "id": "embedding", + "op": "aten.embedding.default", + "args": [ + { + "tensor": "p_model_model_language_model_embed_tokens_weight" + }, + { + "tensor": "input_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1527, in forward\n inputs_embeds = self.get_input_embeddings()(input_ids)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "position_ids" + }, + 0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1179, in forward\n position_ids = position_ids[None, ...].expand(4, position_ids.shape[0], -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze" + }, + [ + 4, + 1, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1179, in forward\n position_ids = position_ids[None, ...].expand(4, position_ids.shape[0], -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_1", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "expand" + }, + 0, + 1, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1183, in forward\n position_ids = position_ids[1:]\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_1", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "b_model_model_language_model_rotary_emb_inv_freq" + }, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 5, in forward\n unsqueeze_1 = torch.ops.aten.unsqueeze.default(b_model_model_language_model_rotary_emb_inv_freq, 0); b_model_model_language_model_rotary_emb_inv_freq = None" + }, + { + "id": "unsqueeze_2", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "unsqueeze_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \".49\", line 6, in forward\n unsqueeze_2 = torch.ops.aten.unsqueeze.default(unsqueeze_1, 1); unsqueeze_1 = None" + }, + { + "id": "unsqueeze_3", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "unsqueeze_2" + }, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 7, in forward\n unsqueeze_3 = torch.ops.aten.unsqueeze.default(unsqueeze_2, 3); unsqueeze_2 = None" + }, + { + "id": "expand_1", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_3" + }, + [ + 3, + 1, + -1, + 1 + ] + ], + "kwargs": {}, + "source": "File \".49\", line 10, in forward\n expand_1 = torch.ops.aten.expand.default(to, [3, 1, -1, 1]); to = None" + }, + { + "id": "unsqueeze_4", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "slice_1" + }, + 2 + ], + "kwargs": {}, + "source": "File \".49\", line 13, in forward\n unsqueeze_4 = torch.ops.aten.unsqueeze.default(slice_1, 2); slice_1 = None" + }, + { + "id": "_to_copy", + "op": "aten._to_copy.default", + "args": [ + { + "tensor": "unsqueeze_4" + } + ], + "kwargs": { + "dtype": { + "dtype": "float32" + } + }, + "source": "File \".49\", line 15, in forward\n to_2 = torch.ops.aten.to.dtype(unsqueeze_4, torch.float32); unsqueeze_4 = None" + }, + { + "id": "matmul", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "expand_1" + }, + { + "tensor": "_to_copy" + } + ], + "kwargs": {}, + "source": "File \".49\", line 20, in forward\n matmul = torch.ops.aten.matmul.default(to_3, to_4); to_3 = to_4 = None" + }, + { + "id": "transpose", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 21, in forward\n transpose = torch.ops.aten.transpose.int(matmul, 2, 3); matmul = None" + }, + { + "id": "select", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 22, in forward\n select = torch.ops.aten.select.int(transpose, 0, 0)" + }, + { + "id": "select_1", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose" + }, + 0, + 1 + ], + "kwargs": {}, + "source": "File \".49\", line 23, in forward\n select_1 = torch.ops.aten.select.int(transpose, 0, 1)" + }, + { + "id": "slice_2", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "select_1" + }, + 2, + 1, + 33, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 24, in forward\n slice_2 = torch.ops.aten.slice.Tensor(select_1, 2, 1, 33, 3); select_1 = None" + }, + { + "id": "slice_3", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "select" + }, + 2, + 1, + 33, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 25, in forward\n slice_3 = torch.ops.aten.slice.Tensor(select, 2, 1, 33, 3)" + }, + { + "id": "copy", + "op": "aten.copy.default", + "args": [ + { + "tensor": "slice_3" + }, + { + "tensor": "slice_2" + } + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "transpose_1", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "select_2", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_1" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "slice_scatter", + "op": "aten.slice_scatter.default", + "args": [ + { + "tensor": "select_2" + }, + { + "tensor": "copy" + }, + 2, + 1, + 33, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "select_scatter", + "op": "aten.select_scatter.default", + "args": [ + { + "tensor": "transpose_1" + }, + { + "tensor": "slice_scatter" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "transpose_2", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "select_scatter" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 26, in forward\n copy_ = torch.ops.aten.copy_.default(slice_3, slice_2); slice_3 = slice_2 = copy_ = None" + }, + { + "id": "transpose_7", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "transpose_2" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 29, in forward\n slice_5 = torch.ops.aten.slice.Tensor(select, 2, 2, 30, 3)" + }, + { + "id": "select_7", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_7" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 29, in forward\n slice_5 = torch.ops.aten.slice.Tensor(select, 2, 2, 30, 3)" + }, + { + "id": "slice_7", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "select_7" + }, + 2, + 2, + 30, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 29, in forward\n slice_5 = torch.ops.aten.slice.Tensor(select, 2, 2, 30, 3)" + }, + { + "id": "transpose_8", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "transpose_2" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 28, in forward\n slice_4 = torch.ops.aten.slice.Tensor(select_2, 2, 2, 30, 3); select_2 = None" + }, + { + "id": "select_8", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_8" + }, + 0, + 2 + ], + "kwargs": {}, + "source": "File \".49\", line 28, in forward\n slice_4 = torch.ops.aten.slice.Tensor(select_2, 2, 2, 30, 3); select_2 = None" + }, + { + "id": "slice_8", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "select_8" + }, + 2, + 2, + 30, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 28, in forward\n slice_4 = torch.ops.aten.slice.Tensor(select_2, 2, 2, 30, 3); select_2 = None" + }, + { + "id": "copy_1", + "op": "aten.copy.default", + "args": [ + { + "tensor": "slice_7" + }, + { + "tensor": "slice_8" + } + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "transpose_9", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "transpose_2" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "select_9", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_9" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "slice_scatter_1", + "op": "aten.slice_scatter.default", + "args": [ + { + "tensor": "select_9" + }, + { + "tensor": "copy_1" + }, + 2, + 2, + 30, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "select_scatter_1", + "op": "aten.select_scatter.default", + "args": [ + { + "tensor": "transpose_9" + }, + { + "tensor": "slice_scatter_1" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "transpose_10", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "select_scatter_1" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 30, in forward\n copy__1 = torch.ops.aten.copy_.default(slice_5, slice_4); slice_5 = slice_4 = copy__1 = None" + }, + { + "id": "transpose_12", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "transpose_10" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \".49\", line 22, in forward\n select = torch.ops.aten.select.int(transpose, 0, 0)" + }, + { + "id": "select_11", + "op": "aten.select.int", + "args": [ + { + "tensor": "transpose_12" + }, + 0, + 0 + ], + "kwargs": {}, + "source": "File \".49\", line 22, in forward\n select = torch.ops.aten.select.int(transpose, 0, 0)" + }, + { + "id": "cat", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "select_11" + }, + { + "tensor": "select_11" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \".49\", line 31, in forward\n cat = torch.ops.aten.cat.default([select, select], -1); select = None" + }, + { + "id": "cos", + "op": "aten.cos.default", + "args": [ + { + "tensor": "cat" + } + ], + "kwargs": {}, + "source": "File \".49\", line 32, in forward\n cos = torch.ops.aten.cos.default(cat)" + }, + { + "id": "mul", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cos" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \".49\", line 33, in forward\n mul = torch.ops.aten.mul.Tensor(cos, 1.0); cos = None" + }, + { + "id": "sin", + "op": "aten.sin.default", + "args": [ + { + "tensor": "cat" + } + ], + "kwargs": {}, + "source": "File \".49\", line 34, in forward\n sin = torch.ops.aten.sin.default(cat); cat = None" + }, + { + "id": "mul_1", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "sin" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \".49\", line 35, in forward\n mul_1 = torch.ops.aten.mul.Tensor(sin, 1.0); sin = None" + }, + { + "id": "pow_1", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "embedding" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_1" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_2", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "embedding" + }, + { + "tensor": "rsqrt" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_1", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_3", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_2" + }, + { + "tensor": "add_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_3" + }, + { + "tensor": "embedding" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_13", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_1", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_1" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_2", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_3", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_1", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_0" + }, + { + "tensor": "transpose_13" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_10", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_1" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_5", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_1" + }, + { + "tensor": "unsqueeze_5" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_11", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_12", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_14", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_12" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_13", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_14" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_14", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_14" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_15", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_14" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_1", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_13" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_2", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_14" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_3", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_15" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_2", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_3" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_4", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg" + }, + { + "tensor": "softplus" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_5", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_1" + }, + { + "tensor": "view_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_1", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_5" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_3", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_1" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_1", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_6", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_1" + }, + { + "tensor": "rsqrt_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_7", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_2" + }, + { + "tensor": "view_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_2", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_7" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_4", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_2" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_2", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_8", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_2" + }, + { + "tensor": "rsqrt_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_6" + }, + { + "tensor": "mul_8" + }, + { + "tensor": "view_3" + }, + { + "tensor": "mul_4" + }, + { + "tensor": "sigmoid" + }, + { + "tensor": "states_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_16", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_17", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_15", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_17" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_15" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_4", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_5", + "op": "aten.view.default", + "args": [ + { + "tensor": "view" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_2", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_4" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_1", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_2" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_5", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_1" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_3", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_9", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_4" + }, + { + "tensor": "rsqrt_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_10", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_norm_weight" + }, + { + "tensor": "mul_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_1", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_11", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_10" + }, + { + "tensor": "silu_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_6", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_11" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_4", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_6" + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_6", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "embedding" + }, + { + "tensor": "linear_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_3", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_6" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_2", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_3" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_7", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_2" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_4", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_12", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_6" + }, + { + "tensor": "rsqrt_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_8", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_0_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_13", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_12" + }, + { + "tensor": "add_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_1", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_13" + }, + { + "tensor": "add_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_5", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_1" + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_2", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_6", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_1" + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_14", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_2" + }, + { + "tensor": "linear_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_7", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_14" + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_9", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_6" + }, + { + "tensor": "linear_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_4", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_9" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_3", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_4" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_10", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_3" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_5", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_15", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_9" + }, + { + "tensor": "rsqrt_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_11", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_16", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_15" + }, + { + "tensor": "add_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_2", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_16" + }, + { + "tensor": "add_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_8", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_2" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_16", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_8" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_9", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_2" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_7", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_9" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_10", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_2" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_11", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_2" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_2", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_2" + }, + { + "tensor": "transpose_16" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_18", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_2" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_1", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_6", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_1", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_2" + }, + { + "tensor": "unsqueeze_6" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_19", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_1" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_3", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_20", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_3" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_17", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_20" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_21", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_17" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_22", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_17" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_23", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_17" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_8", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_21" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_9", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_22" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_10", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_23" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_1", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_1", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_1", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_12", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_11" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_1", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_17", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_1" + }, + { + "tensor": "softplus_1" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_18", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_8" + }, + { + "tensor": "view_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_3", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_18" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_13", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_3" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_6", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_19", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_8" + }, + { + "tensor": "rsqrt_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_20", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_9" + }, + { + "tensor": "view_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_4", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_20" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_14", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_4" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_7", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_21", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_9" + }, + { + "tensor": "rsqrt_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_1", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_19" + }, + { + "tensor": "mul_21" + }, + { + "tensor": "view_10" + }, + { + "tensor": "mul_17" + }, + { + "tensor": "sigmoid_1" + }, + { + "tensor": "states_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_24", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_1" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_25", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_1" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_18", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_25" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_1", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_18" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_11", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_1" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_12", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_7" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_5", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_11" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_4", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_5" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_15", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_4" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_8", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_22", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_11" + }, + { + "tensor": "rsqrt_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_23", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_norm_weight" + }, + { + "tensor": "mul_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_4", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_24", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_23" + }, + { + "tensor": "silu_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_13", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_24" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_12", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_13" + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_16", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_9" + }, + { + "tensor": "linear_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_6", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_16" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_5", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_6" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_17", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_5" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_9", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_25", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_16" + }, + { + "tensor": "rsqrt_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_18", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_1_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_26", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_25" + }, + { + "tensor": "add_18" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_3", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_26" + }, + { + "tensor": "add_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_13", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_3" + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_5", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_14", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_3" + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_27", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_5" + }, + { + "tensor": "linear_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_15", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_27" + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_19", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_16" + }, + { + "tensor": "linear_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_7", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_19" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_6", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_7" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_20", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_6" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_10", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_20" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_28", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_19" + }, + { + "tensor": "rsqrt_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_21", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_29", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_28" + }, + { + "tensor": "add_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_4", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_29" + }, + { + "tensor": "add_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_16", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_4" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_19", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_16" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_17", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_4" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_14", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_17" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_18", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_4" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_19", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_4" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_3", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_4" + }, + { + "tensor": "transpose_19" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_26", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_3" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_2", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_7", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_2" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_2", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_3" + }, + { + "tensor": "unsqueeze_7" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_27", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_2" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_6", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_28", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_6" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_20", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_28" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_29", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_20" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_30", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_20" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_31", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_20" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_15", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_29" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_16", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_30" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_17", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_31" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_2", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_18" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_2", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_2", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_22", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_19" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_2", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_30", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_2" + }, + { + "tensor": "softplus_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_31", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_15" + }, + { + "tensor": "view_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_5", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_31" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_23", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_5" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_11", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_32", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_15" + }, + { + "tensor": "rsqrt_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_33", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_16" + }, + { + "tensor": "view_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_6", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_33" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_24", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_6" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_12", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_24" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_34", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_16" + }, + { + "tensor": "rsqrt_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_2", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_32" + }, + { + "tensor": "mul_34" + }, + { + "tensor": "view_17" + }, + { + "tensor": "mul_30" + }, + { + "tensor": "sigmoid_2" + }, + { + "tensor": "states_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_32", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_2" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_33", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_2" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_21", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_33" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_2", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_21" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_18", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_2" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_19", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_14" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_8", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_18" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_7", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_8" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_25", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_7" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_13", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_25" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_35", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_18" + }, + { + "tensor": "rsqrt_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_36", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_norm_weight" + }, + { + "tensor": "mul_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_7", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_37", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_36" + }, + { + "tensor": "silu_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_20", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_37" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_20", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_20" + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_26", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_19" + }, + { + "tensor": "linear_20" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_9", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_26" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_8", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_9" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_27", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_8" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_14", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_38", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_26" + }, + { + "tensor": "rsqrt_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_28", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_2_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_39", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_38" + }, + { + "tensor": "add_28" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_5", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_39" + }, + { + "tensor": "add_26" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_21", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_5" + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_8", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_22", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_5" + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_40", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_8" + }, + { + "tensor": "linear_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_23", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_40" + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_29", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_26" + }, + { + "tensor": "linear_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_10", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_29" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_9", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_10" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_30", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_9" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_15", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_41", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_29" + }, + { + "tensor": "rsqrt_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_31", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_42", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_41" + }, + { + "tensor": "add_31" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_6", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_42" + }, + { + "tensor": "add_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_24", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_6" + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_21", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_24" + }, + [ + 1, + 64, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_34", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_21" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_35", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_21" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_3", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_35" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_3" + }, + [ + 1, + 64, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_22", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_34" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_11", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_22" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_10", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_11" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_32", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_10" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_16", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_43", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_22" + }, + { + "tensor": "rsqrt_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_33", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_44", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_43" + }, + { + "tensor": "add_33" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_7", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_44" + }, + { + "tensor": "view_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_22", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_7" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_25", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_6" + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_23", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_25" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_12", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_23" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_11", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_12" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_34", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_11" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_17", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_34" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_45", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_23" + }, + { + "tensor": "rsqrt_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_35", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_46", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_45" + }, + { + "tensor": "add_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_8", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_46" + }, + { + "tensor": "view_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_23", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_8" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_26", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_6" + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_24", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_26" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_24", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_24" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_8", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_9", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_36", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_22" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_37", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_22" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_38", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_23" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_39", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_23" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_47", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_36" + }, + { + "tensor": "unsqueeze_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_40", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_36" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_41", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_36" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_3", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_4", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_3" + }, + { + "tensor": "slice_40" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_48", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_4" + }, + { + "tensor": "unsqueeze_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_36", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_47" + }, + { + "tensor": "mul_48" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_49", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_38" + }, + { + "tensor": "unsqueeze_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_42", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_38" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_43", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_38" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_4", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_5", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_4" + }, + { + "tensor": "slice_42" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_50", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_5" + }, + { + "tensor": "unsqueeze_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_37", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_49" + }, + { + "tensor": "mul_50" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_6", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_36" + }, + { + "tensor": "slice_37" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_7", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_37" + }, + { + "tensor": "slice_39" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_6" + }, + { + "tensor": "cat_7" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_1", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_7" + }, + { + "tensor": "transpose_24" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_10", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_2", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_10" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_4", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_2" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_1", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_4" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_11", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_1" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_3", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_11" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_5", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_3" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_2", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_5" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_25", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_1" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_1", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_6" + }, + { + "tensor": "transpose_25" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_51", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_1" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_38", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_51" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_38" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_2", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax" + }, + { + "tensor": "_unsafe_view_2" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_26", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_2" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_6", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_26" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_25", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_6" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_3", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_52", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_25" + }, + { + "tensor": "sigmoid_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_27", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_52" + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_39", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_29" + }, + { + "tensor": "linear_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_13", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_39" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_12", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_13" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_40", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_12" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_18", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_40" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_53", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_39" + }, + { + "tensor": "rsqrt_18" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_41", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_3_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_54", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_53" + }, + { + "tensor": "add_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_9", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_54" + }, + { + "tensor": "add_39" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_28", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_9" + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_9", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_28" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_29", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_9" + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_55", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_9" + }, + { + "tensor": "linear_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_30", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_55" + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_42", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_39" + }, + { + "tensor": "linear_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_14", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_42" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_13", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_14" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_43", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_13" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_19", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_56", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_42" + }, + { + "tensor": "rsqrt_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_44", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_57", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_56" + }, + { + "tensor": "add_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_10", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_57" + }, + { + "tensor": "add_42" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_31", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_10" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_27", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_31" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_32", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_10" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_26", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_32" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_33", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_10" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_34", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_10" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_8", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_8" + }, + { + "tensor": "transpose_27" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_44", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_8" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_3", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_12", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_3" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_3", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_8" + }, + { + "tensor": "unsqueeze_12" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_45", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_3" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_10", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_46", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_10" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_28", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_46" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_47", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_28" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_48", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_28" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_49", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_28" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_27", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_47" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_28", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_48" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_29", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_49" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_4", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_33" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_3", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_5", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_45", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_34" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_3", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_58", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_5" + }, + { + "tensor": "softplus_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_59", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_27" + }, + { + "tensor": "view_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_7", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_59" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_46", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_7" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_20", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_46" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_60", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_27" + }, + { + "tensor": "rsqrt_20" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_61", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_28" + }, + { + "tensor": "view_28" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_8", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_61" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_47", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_8" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_21", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_47" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_62", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_28" + }, + { + "tensor": "rsqrt_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_3", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_60" + }, + { + "tensor": "mul_62" + }, + { + "tensor": "view_29" + }, + { + "tensor": "mul_58" + }, + { + "tensor": "sigmoid_4" + }, + { + "tensor": "states_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_50", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_3" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_51", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_3" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_29", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_51" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_7", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_29" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_30", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_7" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_31", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_26" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_15", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_30" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_14", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_15" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_48", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_14" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_22", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_48" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_63", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_30" + }, + { + "tensor": "rsqrt_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_64", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_norm_weight" + }, + { + "tensor": "mul_63" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_11", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_31" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_65", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_64" + }, + { + "tensor": "silu_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_32", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_65" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_35", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_32" + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_49", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_42" + }, + { + "tensor": "linear_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_16", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_49" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_15", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_16" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_50", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_15" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_23", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_50" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_66", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_49" + }, + { + "tensor": "rsqrt_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_51", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_4_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_67", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_66" + }, + { + "tensor": "add_51" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_11", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_67" + }, + { + "tensor": "add_49" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_36", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_11" + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_12", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_36" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_37", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_11" + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_68", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_12" + }, + { + "tensor": "linear_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_38", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_68" + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_52", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_49" + }, + { + "tensor": "linear_38" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_17", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_52" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_16", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_17" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_53", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_16" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_24", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_69", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_52" + }, + { + "tensor": "rsqrt_24" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_54", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_70", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_69" + }, + { + "tensor": "add_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_12", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_70" + }, + { + "tensor": "add_52" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_39", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_12" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_30", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_39" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_40", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_12" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_33", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_40" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_41", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_12" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_42", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_12" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_9", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_10" + }, + { + "tensor": "transpose_30" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_52", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_9" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_4", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_13", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_4" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_4", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_9" + }, + { + "tensor": "unsqueeze_13" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_53", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_4" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_13", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_54", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_13" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_31", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_54" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_55", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_31" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_56", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_31" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_57", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_31" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_34", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_55" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_35", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_56" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_36", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_57" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_5", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_4", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_6", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_55", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_42" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_4", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_55" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_71", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_6" + }, + { + "tensor": "softplus_4" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_72", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_34" + }, + { + "tensor": "view_34" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_9", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_72" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_56", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_9" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_25", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_56" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_73", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_34" + }, + { + "tensor": "rsqrt_25" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_74", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_35" + }, + { + "tensor": "view_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_10", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_74" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_57", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_10" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_26", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_57" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_75", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_35" + }, + { + "tensor": "rsqrt_26" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_4", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_73" + }, + { + "tensor": "mul_75" + }, + { + "tensor": "view_36" + }, + { + "tensor": "mul_71" + }, + { + "tensor": "sigmoid_5" + }, + { + "tensor": "states_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_58", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_4" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_59", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_4" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_32", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_59" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_8", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_32" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_37", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_8" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_38", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_33" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_18", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_37" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_17", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_18" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_58", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_17" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_27", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_58" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_76", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_37" + }, + { + "tensor": "rsqrt_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_77", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_norm_weight" + }, + { + "tensor": "mul_76" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_14", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_38" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_78", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_77" + }, + { + "tensor": "silu_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_39", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_78" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_43", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_39" + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_59", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_52" + }, + { + "tensor": "linear_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_19", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_59" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_18", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_19" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_60", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_18" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_28", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_79", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_59" + }, + { + "tensor": "rsqrt_28" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_61", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_5_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_80", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_79" + }, + { + "tensor": "add_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_13", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_80" + }, + { + "tensor": "add_59" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_44", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_13" + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_15", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_45", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_13" + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_81", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_15" + }, + { + "tensor": "linear_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_46", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_81" + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_62", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_59" + }, + { + "tensor": "linear_46" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_20", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_62" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_19", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_20" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_63", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_19" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_29", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_63" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_82", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_62" + }, + { + "tensor": "rsqrt_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_64", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_83", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_82" + }, + { + "tensor": "add_64" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_14", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_83" + }, + { + "tensor": "add_62" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_47", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_14" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_33", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_47" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_48", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_14" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_40", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_48" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_49", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_14" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_50", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_14" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_10", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_12" + }, + { + "tensor": "transpose_33" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_60", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_10" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_5", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_14", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_5" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_5", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_10" + }, + { + "tensor": "unsqueeze_14" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_61", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_5" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_16", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_62", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_16" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_34", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_62" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_63", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_34" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_64", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_34" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_65", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_34" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_41", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_63" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_42", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_64" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_43", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_65" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_6", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_49" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_5", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_7", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_65", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_50" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_5", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_65" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_84", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_7" + }, + { + "tensor": "softplus_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_85", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_41" + }, + { + "tensor": "view_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_11", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_85" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_66", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_11" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_30", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_66" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_86", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_41" + }, + { + "tensor": "rsqrt_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_87", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_42" + }, + { + "tensor": "view_42" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_12", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_87" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_67", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_12" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_31", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_88", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_42" + }, + { + "tensor": "rsqrt_31" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_5", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_86" + }, + { + "tensor": "mul_88" + }, + { + "tensor": "view_43" + }, + { + "tensor": "mul_84" + }, + { + "tensor": "sigmoid_6" + }, + { + "tensor": "states_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_66", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_5" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_67", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_5" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_35", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_67" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_9", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_35" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_44", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_9" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_45", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_40" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_21", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_44" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_20", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_21" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_68", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_20" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_32", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_68" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_89", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_44" + }, + { + "tensor": "rsqrt_32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_90", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_norm_weight" + }, + { + "tensor": "mul_89" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_17", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_91", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_90" + }, + { + "tensor": "silu_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_46", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_91" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_51", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_46" + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_69", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_62" + }, + { + "tensor": "linear_51" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_22", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_69" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_21", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_22" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_70", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_21" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_33", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_70" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_92", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_69" + }, + { + "tensor": "rsqrt_33" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_71", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_6_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_93", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_92" + }, + { + "tensor": "add_71" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_15", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_93" + }, + { + "tensor": "add_69" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_52", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_15" + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_18", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_52" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_53", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_15" + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_94", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_18" + }, + { + "tensor": "linear_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_54", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_94" + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_72", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_69" + }, + { + "tensor": "linear_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_23", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_72" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_22", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_23" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_73", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_22" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_34", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_73" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_95", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_72" + }, + { + "tensor": "rsqrt_34" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_74", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_96", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_95" + }, + { + "tensor": "add_74" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_16", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_96" + }, + { + "tensor": "add_72" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_55", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_16" + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_47", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_55" + }, + [ + 1, + 64, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_68", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_47" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_69", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_47" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_10", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_69" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_3", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_10" + }, + [ + 1, + 64, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_48", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_68" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_24", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_48" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_23", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_24" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_75", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_23" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_35", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_97", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_48" + }, + { + "tensor": "rsqrt_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_76", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_98", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_97" + }, + { + "tensor": "add_76" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_17", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_98" + }, + { + "tensor": "view_48" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_36", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_17" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_56", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_16" + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_49", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_56" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_25", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_49" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_24", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_25" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_77", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_24" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_36", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_77" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_99", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_49" + }, + { + "tensor": "rsqrt_36" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_78", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_100", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_99" + }, + { + "tensor": "add_78" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_18", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_100" + }, + { + "tensor": "view_49" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_37", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_18" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_57", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_16" + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_50", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_57" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_38", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_50" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_15", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_16", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_70", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_36" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_71", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_36" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_72", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_37" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_73", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_37" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_101", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_70" + }, + { + "tensor": "unsqueeze_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_74", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_70" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_75", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_70" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_8", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_11", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_8" + }, + { + "tensor": "slice_74" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_102", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_11" + }, + { + "tensor": "unsqueeze_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_79", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_101" + }, + { + "tensor": "mul_102" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_103", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_72" + }, + { + "tensor": "unsqueeze_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_76", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_72" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_77", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_72" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_9", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_77" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_12", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_9" + }, + { + "tensor": "slice_76" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_104", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_12" + }, + { + "tensor": "unsqueeze_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_80", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_103" + }, + { + "tensor": "mul_104" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_13", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_79" + }, + { + "tensor": "slice_71" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_14", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_80" + }, + { + "tensor": "slice_73" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_2", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_14" + }, + { + "tensor": "cat_14" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_3", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_15" + }, + { + "tensor": "transpose_38" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_17", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_2" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_4", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_17" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_11", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_4" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_4", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_11" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_18", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_3" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_5", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_18" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_12", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_5" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_5", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_12" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_39", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_4" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_3", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_13" + }, + { + "tensor": "transpose_39" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_105", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_3" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_81", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_105" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_1", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_81" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_4", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_1" + }, + { + "tensor": "_unsafe_view_5" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_40", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_4" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_13", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_40" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_51", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_13" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_7", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_3" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_106", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_51" + }, + { + "tensor": "sigmoid_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_58", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_106" + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_82", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_72" + }, + { + "tensor": "linear_58" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_26", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_82" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_25", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_26" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_83", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_25" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_37", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_83" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_107", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_82" + }, + { + "tensor": "rsqrt_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_84", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_7_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_108", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_107" + }, + { + "tensor": "add_84" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_19", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_108" + }, + { + "tensor": "add_82" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_59", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_19" + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_19", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_59" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_60", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_19" + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_109", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_19" + }, + { + "tensor": "linear_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_61", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_109" + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_85", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_82" + }, + { + "tensor": "linear_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_27", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_85" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_26", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_27" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_86", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_26" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_38", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_86" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_110", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_85" + }, + { + "tensor": "rsqrt_38" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_87", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_111", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_110" + }, + { + "tensor": "add_87" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_20", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_111" + }, + { + "tensor": "add_85" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_62", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_20" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_41", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_62" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_63", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_20" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_52", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_63" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_64", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_20" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_65", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_20" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_15", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_16" + }, + { + "tensor": "transpose_41" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_78", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_15" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_6", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_19", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_6" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_6", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_15" + }, + { + "tensor": "unsqueeze_19" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_79", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_6" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_20", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_79" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_80", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_20" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_42", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_80" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_81", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_42" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_82", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_42" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_83", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_42" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_53", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_81" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_54", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_82" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_55", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_83" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_8", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_64" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_6", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_10", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_88", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_65" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_6", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_88" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_112", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_10" + }, + { + "tensor": "softplus_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_113", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_53" + }, + { + "tensor": "view_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_13", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_113" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_89", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_13" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_39", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_89" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_114", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_53" + }, + { + "tensor": "rsqrt_39" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_115", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_54" + }, + { + "tensor": "view_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_14", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_115" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_90", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_14" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_40", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_90" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_116", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_54" + }, + { + "tensor": "rsqrt_40" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_6", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_114" + }, + { + "tensor": "mul_116" + }, + { + "tensor": "view_55" + }, + { + "tensor": "mul_112" + }, + { + "tensor": "sigmoid_8" + }, + { + "tensor": "states_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_84", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_6" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_85", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_6" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_43", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_85" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_14", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_43" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_56", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_14" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_57", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_52" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_28", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_56" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_27", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_28" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_91", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_27" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_41", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_91" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_117", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_56" + }, + { + "tensor": "rsqrt_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_118", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_norm_weight" + }, + { + "tensor": "mul_117" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_21", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_57" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_119", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_118" + }, + { + "tensor": "silu_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_58", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_119" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_66", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_58" + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_92", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_85" + }, + { + "tensor": "linear_66" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_29", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_92" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_28", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_29" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_93", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_28" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_42", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_93" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_120", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_92" + }, + { + "tensor": "rsqrt_42" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_94", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_8_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_121", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_120" + }, + { + "tensor": "add_94" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_21", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_121" + }, + { + "tensor": "add_92" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_67", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_21" + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_22", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_68", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_21" + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_122", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_22" + }, + { + "tensor": "linear_68" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_69", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_122" + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_95", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_92" + }, + { + "tensor": "linear_69" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_30", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_95" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_29", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_30" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_96", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_29" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_43", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_96" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_123", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_95" + }, + { + "tensor": "rsqrt_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_97", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_124", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_123" + }, + { + "tensor": "add_97" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_22", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_124" + }, + { + "tensor": "add_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_70", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_22" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_44", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_70" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_71", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_22" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_59", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_71" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_72", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_22" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_73", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_22" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_16", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_18" + }, + { + "tensor": "transpose_44" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_86", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_16" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_7", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_20", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_7" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_7", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_16" + }, + { + "tensor": "unsqueeze_20" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_87", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_7" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_23", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_87" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_88", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_23" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_45", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_88" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_89", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_45" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_90", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_45" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_91", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_45" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_60", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_89" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_61", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_90" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_62", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_91" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_9", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_72" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_7", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_11", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_98", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_73" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_7", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_98" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_125", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_11" + }, + { + "tensor": "softplus_7" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_126", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_60" + }, + { + "tensor": "view_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_15", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_126" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_99", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_15" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_44", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_99" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_127", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_60" + }, + { + "tensor": "rsqrt_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_128", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_61" + }, + { + "tensor": "view_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_16", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_128" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_100", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_16" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_45", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_100" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_129", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_61" + }, + { + "tensor": "rsqrt_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_7", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_127" + }, + { + "tensor": "mul_129" + }, + { + "tensor": "view_62" + }, + { + "tensor": "mul_125" + }, + { + "tensor": "sigmoid_9" + }, + { + "tensor": "states_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_92", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_7" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_93", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_7" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_46", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_93" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_15", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_46" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_63", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_15" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_64", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_59" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_31", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_63" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_30", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_31" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_101", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_30" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_46", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_101" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_130", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_63" + }, + { + "tensor": "rsqrt_46" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_131", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_norm_weight" + }, + { + "tensor": "mul_130" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_24", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_64" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_132", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_131" + }, + { + "tensor": "silu_24" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_65", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_132" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_74", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_65" + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_102", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_95" + }, + { + "tensor": "linear_74" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_32", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_102" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_31", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_32" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_103", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_31" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_47", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_103" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_133", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_102" + }, + { + "tensor": "rsqrt_47" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_104", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_9_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_134", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_133" + }, + { + "tensor": "add_104" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_23", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_134" + }, + { + "tensor": "add_102" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_75", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_23" + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_25", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_76", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_23" + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_135", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_25" + }, + { + "tensor": "linear_76" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_77", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_135" + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_105", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_102" + }, + { + "tensor": "linear_77" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_33", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_105" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_32", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_33" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_106", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_32" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_48", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_106" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_136", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_105" + }, + { + "tensor": "rsqrt_48" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_107", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_137", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_136" + }, + { + "tensor": "add_107" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_24", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_137" + }, + { + "tensor": "add_105" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_78", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_24" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_47", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_78" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_79", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_24" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_66", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_79" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_80", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_24" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_81", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_24" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_17", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_20" + }, + { + "tensor": "transpose_47" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_94", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_17" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_8", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_21", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_8" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_8", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_17" + }, + { + "tensor": "unsqueeze_21" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_95", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_8" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_26", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_96", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_26" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_48", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_96" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_97", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_48" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_98", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_48" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_99", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_48" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_67", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_97" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_68", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_98" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_69", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_99" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_10", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_80" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_8", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_12", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_108", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_81" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_8", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_108" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_138", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_12" + }, + { + "tensor": "softplus_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_139", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_67" + }, + { + "tensor": "view_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_17", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_139" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_109", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_17" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_49", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_109" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_140", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_67" + }, + { + "tensor": "rsqrt_49" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_141", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_68" + }, + { + "tensor": "view_68" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_18", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_141" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_110", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_18" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_50", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_110" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_142", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_68" + }, + { + "tensor": "rsqrt_50" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_8", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_140" + }, + { + "tensor": "mul_142" + }, + { + "tensor": "view_69" + }, + { + "tensor": "mul_138" + }, + { + "tensor": "sigmoid_10" + }, + { + "tensor": "states_21" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_100", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_8" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_101", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_8" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_49", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_101" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_16", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_49" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_70", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_16" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_71", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_66" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_34", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_70" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_33", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_34" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_111", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_33" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_51", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_111" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_143", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_70" + }, + { + "tensor": "rsqrt_51" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_144", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_norm_weight" + }, + { + "tensor": "mul_143" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_27", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_71" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_145", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_144" + }, + { + "tensor": "silu_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_72", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_145" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_82", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_72" + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_112", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_105" + }, + { + "tensor": "linear_82" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_35", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_112" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_34", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_35" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_113", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_34" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_52", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_146", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_112" + }, + { + "tensor": "rsqrt_52" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_114", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_10_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_147", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_146" + }, + { + "tensor": "add_114" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_25", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_147" + }, + { + "tensor": "add_112" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_83", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_25" + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_28", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_83" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_84", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_25" + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_148", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_28" + }, + { + "tensor": "linear_84" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_85", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_148" + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_115", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_112" + }, + { + "tensor": "linear_85" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_36", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_115" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_35", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_36" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_116", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_35" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_53", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_116" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_149", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_115" + }, + { + "tensor": "rsqrt_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_117", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_150", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_149" + }, + { + "tensor": "add_117" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_26", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_150" + }, + { + "tensor": "add_115" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_86", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_26" + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_73", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_86" + }, + [ + 1, + 64, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_102", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_73" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_103", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_73" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_17", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_103" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_6", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_17" + }, + [ + 1, + 64, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_74", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_102" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_37", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_74" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_36", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_37" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_118", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_36" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_54", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_118" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_151", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_74" + }, + { + "tensor": "rsqrt_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_119", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_152", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_151" + }, + { + "tensor": "add_119" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_27", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_152" + }, + { + "tensor": "view_74" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_50", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_27" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_87", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_26" + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_75", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_87" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_38", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_75" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_37", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_38" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_120", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_37" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_55", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_120" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_153", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_75" + }, + { + "tensor": "rsqrt_55" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_121", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_154", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_153" + }, + { + "tensor": "add_121" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_28", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_154" + }, + { + "tensor": "view_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_51", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_28" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_88", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_26" + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_76", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_88" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_52", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_76" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_22", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_23", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_104", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_50" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_105", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_50" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_106", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_51" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_107", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_51" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_155", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_104" + }, + { + "tensor": "unsqueeze_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_108", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_104" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_109", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_104" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_13", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_109" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_18", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_13" + }, + { + "tensor": "slice_108" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_156", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_18" + }, + { + "tensor": "unsqueeze_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_122", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_155" + }, + { + "tensor": "mul_156" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_157", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_106" + }, + { + "tensor": "unsqueeze_22" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_110", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_106" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_111", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_106" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_14", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_111" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_19", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_14" + }, + { + "tensor": "slice_110" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_158", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_19" + }, + { + "tensor": "unsqueeze_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_123", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_157" + }, + { + "tensor": "mul_158" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_20", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_122" + }, + { + "tensor": "slice_105" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_21", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_123" + }, + { + "tensor": "slice_107" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_4", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_22" + }, + { + "tensor": "cat_21" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_5", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_23" + }, + { + "tensor": "transpose_52" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_24", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_4" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_6", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_24" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_18", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_6" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_7", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_18" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_25", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_5" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_7", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_25" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_19", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_7" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_8", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_19" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_53", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_7" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_5", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_20" + }, + { + "tensor": "transpose_53" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_159", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_5" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_124", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_159" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_2", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_124" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_6", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_2" + }, + { + "tensor": "_unsafe_view_8" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_54", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_6" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_20", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_54" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_77", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_20" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_11", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_6" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_160", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_77" + }, + { + "tensor": "sigmoid_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_89", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_160" + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_125", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_115" + }, + { + "tensor": "linear_89" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_39", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_125" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_38", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_39" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_126", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_38" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_56", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_126" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_161", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_125" + }, + { + "tensor": "rsqrt_56" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_127", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_11_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_162", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_161" + }, + { + "tensor": "add_127" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_29", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_162" + }, + { + "tensor": "add_125" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_90", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_29" + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_29", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_90" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_91", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_29" + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_163", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_29" + }, + { + "tensor": "linear_91" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_92", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_163" + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_128", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_125" + }, + { + "tensor": "linear_92" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_40", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_128" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_39", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_40" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_129", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_39" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_57", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_129" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_164", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_128" + }, + { + "tensor": "rsqrt_57" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_130", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_165", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_164" + }, + { + "tensor": "add_130" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_30", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_165" + }, + { + "tensor": "add_128" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_93", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_30" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_55", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_93" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_94", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_30" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_78", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_94" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_95", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_30" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_96", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_30" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_22", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_24" + }, + { + "tensor": "transpose_55" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_112", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_22" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_9", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_26", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_9" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_9", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_22" + }, + { + "tensor": "unsqueeze_26" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_113", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_9" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_30", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_114", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_30" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_56", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_114" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_115", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_56" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_116", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_56" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_117", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_56" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_79", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_115" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_80", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_116" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_81", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_117" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_12", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_9", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_15", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_131", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_96" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_9", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_131" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_166", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_15" + }, + { + "tensor": "softplus_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_167", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_79" + }, + { + "tensor": "view_79" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_19", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_167" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_132", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_19" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_58", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_132" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_168", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_79" + }, + { + "tensor": "rsqrt_58" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_169", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_80" + }, + { + "tensor": "view_80" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_20", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_169" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_133", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_20" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_59", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_133" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_170", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_80" + }, + { + "tensor": "rsqrt_59" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_9", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_168" + }, + { + "tensor": "mul_170" + }, + { + "tensor": "view_81" + }, + { + "tensor": "mul_166" + }, + { + "tensor": "sigmoid_12" + }, + { + "tensor": "states_25" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_118", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_9" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_119", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_9" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_57", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_119" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_21", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_57" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_82", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_21" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_83", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_78" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_41", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_82" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_40", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_41" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_134", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_40" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_60", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_134" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_171", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_82" + }, + { + "tensor": "rsqrt_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_172", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_norm_weight" + }, + { + "tensor": "mul_171" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_31", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_83" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_173", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_172" + }, + { + "tensor": "silu_31" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_84", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_173" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_97", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_84" + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_135", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_128" + }, + { + "tensor": "linear_97" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_42", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_135" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_41", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_42" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_136", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_41" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_61", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_136" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_174", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_135" + }, + { + "tensor": "rsqrt_61" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_137", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_12_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_175", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_174" + }, + { + "tensor": "add_137" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_31", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_175" + }, + { + "tensor": "add_135" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_98", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_31" + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_32", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_98" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_99", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_31" + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_176", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_32" + }, + { + "tensor": "linear_99" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_100", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_176" + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_138", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_135" + }, + { + "tensor": "linear_100" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_43", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_138" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_42", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_43" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_139", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_42" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_62", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_139" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_177", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_138" + }, + { + "tensor": "rsqrt_62" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_140", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_178", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_177" + }, + { + "tensor": "add_140" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_32", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_178" + }, + { + "tensor": "add_138" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_101", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_32" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_58", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_101" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_102", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_32" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_85", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_102" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_103", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_32" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_104", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_32" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_23", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_26" + }, + { + "tensor": "transpose_58" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_120", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_23" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_10", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_27", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_10" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_10", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_23" + }, + { + "tensor": "unsqueeze_27" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_121", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_10" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_33", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_121" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_122", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_33" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_59", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_122" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_123", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_59" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_124", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_59" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_125", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_59" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_86", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_123" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_87", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_124" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_88", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_125" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_13", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_103" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_10", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_16", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_141", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_104" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_10", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_141" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_179", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_16" + }, + { + "tensor": "softplus_10" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_180", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_86" + }, + { + "tensor": "view_86" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_21", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_180" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_142", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_21" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_63", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_142" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_181", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_86" + }, + { + "tensor": "rsqrt_63" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_182", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_87" + }, + { + "tensor": "view_87" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_22", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_182" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_143", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_22" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_64", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_143" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_183", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_87" + }, + { + "tensor": "rsqrt_64" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_10", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_181" + }, + { + "tensor": "mul_183" + }, + { + "tensor": "view_88" + }, + { + "tensor": "mul_179" + }, + { + "tensor": "sigmoid_13" + }, + { + "tensor": "states_27" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_126", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_10" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_127", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_10" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_60", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_127" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_22", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_60" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_89", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_22" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_90", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_85" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_44", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_89" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_43", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_44" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_144", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_43" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_65", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_144" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_184", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_89" + }, + { + "tensor": "rsqrt_65" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_185", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_norm_weight" + }, + { + "tensor": "mul_184" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_34", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_90" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_186", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_185" + }, + { + "tensor": "silu_34" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_91", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_186" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_105", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_91" + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_145", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_138" + }, + { + "tensor": "linear_105" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_45", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_145" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_44", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_45" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_146", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_44" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_66", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_146" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_187", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_145" + }, + { + "tensor": "rsqrt_66" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_147", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_13_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_188", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_187" + }, + { + "tensor": "add_147" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_33", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_188" + }, + { + "tensor": "add_145" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_106", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_33" + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_35", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_106" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_107", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_33" + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_189", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_35" + }, + { + "tensor": "linear_107" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_108", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_189" + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_148", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_145" + }, + { + "tensor": "linear_108" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_46", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_148" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_45", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_46" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_149", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_45" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_67", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_149" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_190", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_148" + }, + { + "tensor": "rsqrt_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_150", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_191", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_190" + }, + { + "tensor": "add_150" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_34", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_191" + }, + { + "tensor": "add_148" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_109", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_34" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_61", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_109" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_110", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_34" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_92", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_110" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_111", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_34" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_112", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_34" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_24", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_28" + }, + { + "tensor": "transpose_61" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_128", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_24" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_11", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_28", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_11" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_11", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_24" + }, + { + "tensor": "unsqueeze_28" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_129", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_11" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_36", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_129" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_130", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_36" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_62", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_130" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_131", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_62" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_132", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_62" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_133", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_62" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_93", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_131" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_94", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_132" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_95", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_133" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_14", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_111" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_11", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_17", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_151", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_112" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_11", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_151" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_192", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_17" + }, + { + "tensor": "softplus_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_193", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_93" + }, + { + "tensor": "view_93" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_23", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_193" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_152", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_23" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_68", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_152" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_194", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_93" + }, + { + "tensor": "rsqrt_68" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_195", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_94" + }, + { + "tensor": "view_94" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_24", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_195" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_153", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_24" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_69", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_153" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_196", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_94" + }, + { + "tensor": "rsqrt_69" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_11", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_194" + }, + { + "tensor": "mul_196" + }, + { + "tensor": "view_95" + }, + { + "tensor": "mul_192" + }, + { + "tensor": "sigmoid_14" + }, + { + "tensor": "states_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_134", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_11" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_135", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_11" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_63", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_135" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_23", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_63" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_96", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_23" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_97", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_92" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_47", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_96" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_46", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_47" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_154", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_46" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_70", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_154" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_197", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_96" + }, + { + "tensor": "rsqrt_70" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_198", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_norm_weight" + }, + { + "tensor": "mul_197" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_37", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_97" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_199", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_198" + }, + { + "tensor": "silu_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_98", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_199" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_113", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_98" + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_155", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_148" + }, + { + "tensor": "linear_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_48", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_155" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_47", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_48" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_156", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_47" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_71", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_156" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_200", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_155" + }, + { + "tensor": "rsqrt_71" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_157", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_14_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_201", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_200" + }, + { + "tensor": "add_157" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_35", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_201" + }, + { + "tensor": "add_155" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_114", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_35" + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_38", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_114" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_115", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_35" + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_202", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_38" + }, + { + "tensor": "linear_115" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_116", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_202" + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_158", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_155" + }, + { + "tensor": "linear_116" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_49", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_158" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_48", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_49" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_159", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_48" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_72", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_159" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_203", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_158" + }, + { + "tensor": "rsqrt_72" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_160", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_204", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_203" + }, + { + "tensor": "add_160" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_36", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_204" + }, + { + "tensor": "add_158" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_117", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_36" + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_99", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_117" + }, + [ + 1, + 64, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_136", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_99" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_137", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_99" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_24", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_137" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_9", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_24" + }, + [ + 1, + 64, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_100", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_136" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_50", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_100" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_49", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_50" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_161", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_49" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_73", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_161" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_205", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_100" + }, + { + "tensor": "rsqrt_73" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_162", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_206", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_205" + }, + { + "tensor": "add_162" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_37", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_206" + }, + { + "tensor": "view_100" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_64", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_37" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_118", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_36" + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_101", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_118" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_51", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_101" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_50", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_51" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_163", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_50" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_74", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_163" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_207", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_101" + }, + { + "tensor": "rsqrt_74" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_164", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_208", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_207" + }, + { + "tensor": "add_164" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_38", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_208" + }, + { + "tensor": "view_101" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_65", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_38" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_119", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_36" + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_102", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_119" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_66", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_102" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_29", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_30", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_138", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_64" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_139", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_64" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_140", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_65" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_141", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_65" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_209", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_138" + }, + { + "tensor": "unsqueeze_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_142", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_138" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_143", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_138" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_18", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_143" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_25", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_18" + }, + { + "tensor": "slice_142" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_210", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_25" + }, + { + "tensor": "unsqueeze_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_165", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_209" + }, + { + "tensor": "mul_210" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_211", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_140" + }, + { + "tensor": "unsqueeze_29" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_144", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_140" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_145", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_140" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_19", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_145" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_26", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_19" + }, + { + "tensor": "slice_144" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_212", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_26" + }, + { + "tensor": "unsqueeze_30" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_166", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_211" + }, + { + "tensor": "mul_212" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_27", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_165" + }, + { + "tensor": "slice_139" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_28", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_166" + }, + { + "tensor": "slice_141" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_6", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_30" + }, + { + "tensor": "cat_28" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_7", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_31" + }, + { + "tensor": "transpose_66" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_31", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_6" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_8", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_31" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_25", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_8" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_10", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_25" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_32", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_7" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_9", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_32" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_26", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_9" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_11", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_26" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_67", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_10" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_7", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_27" + }, + { + "tensor": "transpose_67" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_213", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_7" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_167", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_213" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_3", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_167" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_8", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_3" + }, + { + "tensor": "_unsafe_view_11" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_68", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_8" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_27", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_68" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_103", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_27" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_15", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_9" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_214", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_103" + }, + { + "tensor": "sigmoid_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_120", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_214" + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_168", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_158" + }, + { + "tensor": "linear_120" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_52", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_168" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_51", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_52" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_169", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_51" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_75", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_169" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_215", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_168" + }, + { + "tensor": "rsqrt_75" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_170", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_15_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_216", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_215" + }, + { + "tensor": "add_170" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_39", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_216" + }, + { + "tensor": "add_168" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_121", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_39" + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_39", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_121" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_122", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_39" + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_217", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_39" + }, + { + "tensor": "linear_122" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_123", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_217" + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_171", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_168" + }, + { + "tensor": "linear_123" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_53", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_171" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_52", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_53" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_172", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_52" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_76", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_172" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_218", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_171" + }, + { + "tensor": "rsqrt_76" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_173", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_219", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_218" + }, + { + "tensor": "add_173" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_40", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_219" + }, + { + "tensor": "add_171" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_124", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_40" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_69", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_124" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_125", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_40" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_104", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_125" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_126", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_40" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_127", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_40" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_29", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_32" + }, + { + "tensor": "transpose_69" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_146", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_29" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_12", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_33", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_12" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_12", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_29" + }, + { + "tensor": "unsqueeze_33" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_147", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_12" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_40", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_147" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_148", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_40" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_70", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_148" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_149", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_70" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_150", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_70" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_151", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_70" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_105", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_149" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_106", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_150" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_107", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_151" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_16", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_126" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_12", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_20", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_174", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_127" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_12", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_174" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_220", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_20" + }, + { + "tensor": "softplus_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_221", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_105" + }, + { + "tensor": "view_105" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_25", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_221" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_175", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_25" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_77", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_175" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_222", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_105" + }, + { + "tensor": "rsqrt_77" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_223", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_106" + }, + { + "tensor": "view_106" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_26", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_223" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_176", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_26" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_78", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_176" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_224", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_106" + }, + { + "tensor": "rsqrt_78" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_12", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_222" + }, + { + "tensor": "mul_224" + }, + { + "tensor": "view_107" + }, + { + "tensor": "mul_220" + }, + { + "tensor": "sigmoid_16" + }, + { + "tensor": "states_33" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_152", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_12" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_153", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_12" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_71", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_153" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_28", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_71" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_108", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_28" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_109", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_104" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_54", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_108" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_53", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_54" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_177", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_53" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_79", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_177" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_225", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_108" + }, + { + "tensor": "rsqrt_79" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_226", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_norm_weight" + }, + { + "tensor": "mul_225" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_41", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_109" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_227", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_226" + }, + { + "tensor": "silu_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_110", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_227" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_128", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_110" + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_178", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_171" + }, + { + "tensor": "linear_128" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_55", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_178" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_54", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_55" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_179", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_54" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_80", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_179" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_228", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_178" + }, + { + "tensor": "rsqrt_80" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_180", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_16_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_229", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_228" + }, + { + "tensor": "add_180" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_41", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_229" + }, + { + "tensor": "add_178" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_129", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_41" + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_42", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_129" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_130", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_41" + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_230", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_42" + }, + { + "tensor": "linear_130" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_131", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_230" + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_181", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_178" + }, + { + "tensor": "linear_131" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_56", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_181" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_55", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_56" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_182", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_55" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_81", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_182" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_231", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_181" + }, + { + "tensor": "rsqrt_81" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_183", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_232", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_231" + }, + { + "tensor": "add_183" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_42", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_232" + }, + { + "tensor": "add_181" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_132", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_42" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_72", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_132" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_133", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_42" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_111", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_133" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_134", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_42" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_135", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_42" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_30", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_34" + }, + { + "tensor": "transpose_72" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_154", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_30" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_13", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_34", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_13" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_13", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_30" + }, + { + "tensor": "unsqueeze_34" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_155", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_13" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_43", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_155" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_156", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_43" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_73", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_156" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_157", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_73" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_158", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_73" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_159", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_73" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_112", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_157" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_113", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_158" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_114", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_159" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_17", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_134" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_13", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_21", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_184", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_135" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_13", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_184" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_233", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_21" + }, + { + "tensor": "softplus_13" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_234", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_112" + }, + { + "tensor": "view_112" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_27", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_234" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_185", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_27" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_82", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_185" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_235", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_112" + }, + { + "tensor": "rsqrt_82" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_236", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_113" + }, + { + "tensor": "view_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_28", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_236" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_186", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_28" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_83", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_186" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_237", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_113" + }, + { + "tensor": "rsqrt_83" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_13", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_235" + }, + { + "tensor": "mul_237" + }, + { + "tensor": "view_114" + }, + { + "tensor": "mul_233" + }, + { + "tensor": "sigmoid_17" + }, + { + "tensor": "states_35" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_160", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_13" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_161", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_13" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_74", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_161" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_29", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_74" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_115", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_29" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_116", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_111" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_57", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_115" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_56", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_57" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_187", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_56" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_84", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_187" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_238", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_115" + }, + { + "tensor": "rsqrt_84" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_239", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_norm_weight" + }, + { + "tensor": "mul_238" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_44", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_116" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_240", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_239" + }, + { + "tensor": "silu_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_117", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_240" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_136", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_117" + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_188", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_181" + }, + { + "tensor": "linear_136" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_58", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_188" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_57", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_58" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_189", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_57" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_85", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_189" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_241", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_188" + }, + { + "tensor": "rsqrt_85" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_190", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_17_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_242", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_241" + }, + { + "tensor": "add_190" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_43", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_242" + }, + { + "tensor": "add_188" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_137", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_43" + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_45", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_137" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_138", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_43" + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_243", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_45" + }, + { + "tensor": "linear_138" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_139", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_243" + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_191", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_188" + }, + { + "tensor": "linear_139" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_59", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_191" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_58", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_59" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_192", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_58" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_86", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_192" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_244", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_191" + }, + { + "tensor": "rsqrt_86" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_193", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_245", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_244" + }, + { + "tensor": "add_193" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_44", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_245" + }, + { + "tensor": "add_191" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_140", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_44" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_75", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_140" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_141", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_44" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_118", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_141" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_142", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_44" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_143", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_44" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_31", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_36" + }, + { + "tensor": "transpose_75" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_162", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_31" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_14", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_35", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_14" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_14", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_31" + }, + { + "tensor": "unsqueeze_35" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_163", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_14" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_46", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_163" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_164", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_46" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_76", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_164" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_165", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_76" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_166", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_76" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_167", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_76" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_119", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_165" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_120", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_166" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_121", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_167" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_18", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_142" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_14", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_22", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_194", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_143" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_14", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_194" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_246", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_22" + }, + { + "tensor": "softplus_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_247", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_119" + }, + { + "tensor": "view_119" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_29", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_247" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_195", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_29" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_87", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_195" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_248", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_119" + }, + { + "tensor": "rsqrt_87" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_249", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_120" + }, + { + "tensor": "view_120" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_30", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_249" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_196", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_30" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_88", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_196" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_250", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_120" + }, + { + "tensor": "rsqrt_88" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_14", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_248" + }, + { + "tensor": "mul_250" + }, + { + "tensor": "view_121" + }, + { + "tensor": "mul_246" + }, + { + "tensor": "sigmoid_18" + }, + { + "tensor": "states_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_168", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_14" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_169", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_14" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_77", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_169" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_30", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_77" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_122", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_30" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_123", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_118" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_60", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_122" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_59", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_60" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_197", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_59" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_89", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_197" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_251", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_122" + }, + { + "tensor": "rsqrt_89" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_252", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_norm_weight" + }, + { + "tensor": "mul_251" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_47", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_123" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_253", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_252" + }, + { + "tensor": "silu_47" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_124", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_253" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_144", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_124" + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_198", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_191" + }, + { + "tensor": "linear_144" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_61", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_198" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_60", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_61" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_199", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_60" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_90", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_199" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_254", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_198" + }, + { + "tensor": "rsqrt_90" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_200", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_18_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_255", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_254" + }, + { + "tensor": "add_200" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_45", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_255" + }, + { + "tensor": "add_198" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_145", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_45" + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_48", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_145" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_146", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_45" + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_256", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_48" + }, + { + "tensor": "linear_146" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_147", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_256" + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_201", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_198" + }, + { + "tensor": "linear_147" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_62", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_201" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_61", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_62" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_202", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_61" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_91", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_202" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_257", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_201" + }, + { + "tensor": "rsqrt_91" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_203", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_258", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_257" + }, + { + "tensor": "add_203" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_46", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_258" + }, + { + "tensor": "add_201" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_148", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_46" + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_125", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_148" + }, + [ + 1, + 64, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_170", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_125" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_171", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_125" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_31", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_171" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_12", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_31" + }, + [ + 1, + 64, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_126", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_170" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_63", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_126" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_62", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_63" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_204", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_62" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_92", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_204" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_259", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_126" + }, + { + "tensor": "rsqrt_92" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_205", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_260", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_259" + }, + { + "tensor": "add_205" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_47", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_260" + }, + { + "tensor": "view_126" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_78", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_47" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_149", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_46" + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_127", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_149" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_64", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_127" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_63", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_64" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_206", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_63" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_93", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_206" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_261", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_127" + }, + { + "tensor": "rsqrt_93" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_207", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_262", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_261" + }, + { + "tensor": "add_207" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_48", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_262" + }, + { + "tensor": "view_127" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_79", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_48" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_150", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_46" + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_128", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_150" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_80", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_128" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_36", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_37", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_172", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_78" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_173", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_78" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_174", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_79" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_175", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_79" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_263", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_172" + }, + { + "tensor": "unsqueeze_36" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_176", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_172" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_177", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_172" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_23", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_177" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_32", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_23" + }, + { + "tensor": "slice_176" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_264", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_32" + }, + { + "tensor": "unsqueeze_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_208", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_263" + }, + { + "tensor": "mul_264" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_265", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_174" + }, + { + "tensor": "unsqueeze_36" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_178", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_174" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_179", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_174" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_24", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_179" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_33", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_24" + }, + { + "tensor": "slice_178" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_266", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_33" + }, + { + "tensor": "unsqueeze_37" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_209", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_265" + }, + { + "tensor": "mul_266" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_34", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_208" + }, + { + "tensor": "slice_173" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_35", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_209" + }, + { + "tensor": "slice_175" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_8", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_38" + }, + { + "tensor": "cat_35" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_9", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_39" + }, + { + "tensor": "transpose_80" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_38", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_8" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_10", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_38" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_32", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_10" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_13", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_32" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_39", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_9" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_11", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_39" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_33", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_11" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_14", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_33" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_81", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_13" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_9", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_34" + }, + { + "tensor": "transpose_81" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_267", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_9" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_210", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_267" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_4", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_210" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_10", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_4" + }, + { + "tensor": "_unsafe_view_14" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_82", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_10" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_34", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_82" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_129", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_34" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_19", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_12" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_268", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_129" + }, + { + "tensor": "sigmoid_19" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_151", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_268" + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_211", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_201" + }, + { + "tensor": "linear_151" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_65", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_211" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_64", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_65" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_212", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_64" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_94", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_212" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_269", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_211" + }, + { + "tensor": "rsqrt_94" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_213", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_19_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_270", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_269" + }, + { + "tensor": "add_213" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_49", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_270" + }, + { + "tensor": "add_211" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_152", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_49" + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_49", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_152" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_153", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_49" + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_271", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_49" + }, + { + "tensor": "linear_153" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_154", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_271" + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_214", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_211" + }, + { + "tensor": "linear_154" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_66", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_214" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_65", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_66" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_215", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_65" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_95", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_215" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_272", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_214" + }, + { + "tensor": "rsqrt_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_216", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_273", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_272" + }, + { + "tensor": "add_216" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_50", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_273" + }, + { + "tensor": "add_214" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_155", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_50" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_83", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_155" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_156", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_50" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_130", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_156" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_157", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_50" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_158", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_50" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_36", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_40" + }, + { + "tensor": "transpose_83" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_180", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_36" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_15", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_40", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_15" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_15", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_36" + }, + { + "tensor": "unsqueeze_40" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_181", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_15" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_50", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_181" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_182", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_50" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_84", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_182" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_183", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_84" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_184", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_84" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_185", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_84" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_131", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_183" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_132", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_184" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_133", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_185" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_20", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_157" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_15", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_25", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_217", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_158" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_15", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_217" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_274", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_25" + }, + { + "tensor": "softplus_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_275", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_131" + }, + { + "tensor": "view_131" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_31", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_275" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_218", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_31" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_96", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_218" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_276", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_131" + }, + { + "tensor": "rsqrt_96" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_277", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_132" + }, + { + "tensor": "view_132" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_32", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_277" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_219", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_32" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_97", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_219" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_278", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_132" + }, + { + "tensor": "rsqrt_97" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_15", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_276" + }, + { + "tensor": "mul_278" + }, + { + "tensor": "view_133" + }, + { + "tensor": "mul_274" + }, + { + "tensor": "sigmoid_20" + }, + { + "tensor": "states_41" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_186", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_15" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_187", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_15" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_85", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_187" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_35", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_85" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_134", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_35" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_135", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_130" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_67", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_134" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_66", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_67" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_220", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_66" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_98", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_220" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_279", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_134" + }, + { + "tensor": "rsqrt_98" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_280", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_norm_weight" + }, + { + "tensor": "mul_279" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_51", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_135" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_281", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_280" + }, + { + "tensor": "silu_51" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_136", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_281" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_159", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_136" + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_221", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_214" + }, + { + "tensor": "linear_159" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_68", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_221" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_67", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_68" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_222", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_67" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_99", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_222" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_282", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_221" + }, + { + "tensor": "rsqrt_99" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_223", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_20_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_283", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_282" + }, + { + "tensor": "add_223" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_51", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_283" + }, + { + "tensor": "add_221" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_160", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_51" + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_52", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_160" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_161", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_51" + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_284", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_52" + }, + { + "tensor": "linear_161" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_162", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_284" + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_224", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_221" + }, + { + "tensor": "linear_162" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_69", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_224" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_68", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_69" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_225", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_68" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_100", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_225" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_285", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_224" + }, + { + "tensor": "rsqrt_100" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_226", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_286", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_285" + }, + { + "tensor": "add_226" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_52", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_286" + }, + { + "tensor": "add_224" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_163", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_52" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_86", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_163" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_164", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_52" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_137", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_164" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_165", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_52" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_166", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_52" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_37", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_42" + }, + { + "tensor": "transpose_86" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_188", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_37" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_16", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_41", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_16" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_16", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_37" + }, + { + "tensor": "unsqueeze_41" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_189", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_16" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_53", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_189" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_190", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_53" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_87", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_190" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_191", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_87" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_192", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_87" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_193", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_87" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_138", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_191" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_139", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_192" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_140", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_193" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_21", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_165" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_16", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_26", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_227", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_166" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_16", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_227" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_287", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_26" + }, + { + "tensor": "softplus_16" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_288", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_138" + }, + { + "tensor": "view_138" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_33", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_288" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_228", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_33" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_101", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_228" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_289", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_138" + }, + { + "tensor": "rsqrt_101" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_290", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_139" + }, + { + "tensor": "view_139" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_34", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_290" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_229", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_34" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_102", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_229" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_291", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_139" + }, + { + "tensor": "rsqrt_102" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_16", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_289" + }, + { + "tensor": "mul_291" + }, + { + "tensor": "view_140" + }, + { + "tensor": "mul_287" + }, + { + "tensor": "sigmoid_21" + }, + { + "tensor": "states_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_194", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_16" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_195", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_16" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_88", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_195" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_36", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_88" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_141", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_36" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_142", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_137" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_70", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_141" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_69", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_70" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_230", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_69" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_103", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_230" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_292", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_141" + }, + { + "tensor": "rsqrt_103" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_293", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_norm_weight" + }, + { + "tensor": "mul_292" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_54", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_142" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_294", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_293" + }, + { + "tensor": "silu_54" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_143", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_294" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_167", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_143" + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_231", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_224" + }, + { + "tensor": "linear_167" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_71", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_231" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_70", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_71" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_232", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_70" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_104", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_232" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_295", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_231" + }, + { + "tensor": "rsqrt_104" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_233", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_21_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_296", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_295" + }, + { + "tensor": "add_233" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_53", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_296" + }, + { + "tensor": "add_231" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_168", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_53" + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_55", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_168" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_169", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_53" + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_297", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_55" + }, + { + "tensor": "linear_169" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_170", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_297" + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_234", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_231" + }, + { + "tensor": "linear_170" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_72", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_234" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_71", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_72" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_235", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_71" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_105", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_235" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_298", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_234" + }, + { + "tensor": "rsqrt_105" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_236", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_299", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_298" + }, + { + "tensor": "add_236" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_54", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_299" + }, + { + "tensor": "add_234" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_171", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_54" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 448, in forward\n mixed_qkv = self.in_proj_qkv(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_89", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "linear_171" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 449, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_172", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_54" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 451, in forward\n z = self.in_proj_z(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_144", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_172" + }, + [ + 1, + 64, + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 452, in forward\n z = z.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_173", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_54" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 454, in forward\n b = self.in_proj_b(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_174", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_54" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 455, in forward\n a = self.in_proj_a(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_38", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "states_44" + }, + { + "tensor": "transpose_89" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 96, in update_conv_state\n combined = torch.cat([self.states[2 * layer_idx], value], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_196", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "cat_38" + }, + 2, + -4, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 469, in forward\n mixed_qkv = cache_params.update_conv_state(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 97, in update_conv_state\n self.states[2 * layer_idx] = combined[..., -conv_kernel_size:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "squeeze_17", + "op": "aten.squeeze.dim", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_conv1d_weight" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 475, in forward\n self.conv1d.weight.squeeze(1)," + }, + { + "id": "unsqueeze_42", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "squeeze_17" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 232, in causal_conv1d_fn\n weight=weight.unsqueeze(1),\n ^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "conv1d_17", + "op": "aten.conv1d.default", + "args": [ + { + "tensor": "cat_38" + }, + { + "tensor": "unsqueeze_42" + }, + null, + [ + 1 + ], + [ + 3 + ], + [ + 1 + ], + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "slice_197", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "conv1d_17" + }, + 2, + 0, + 68 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 230, in causal_conv1d_fn\n out = F.conv1d(\n ^^^^^^^^^" + }, + { + "id": "silu_56", + "op": "aten.silu.default", + "args": [ + { + "tensor": "slice_197" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 473, in forward\n mixed_qkv = causal_conv1d_fn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/hub_kernels.py\", line 861, in wrapped\n return implementation(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 238, in causal_conv1d_fn\n out = ACT2FN[activation](out)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_198", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "silu_56" + }, + 2, + -64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 483, in forward\n mixed_qkv = mixed_qkv[:, :, -seq_len:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_90", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_198" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 485, in forward\n mixed_qkv = mixed_qkv.transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_199", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_90" + }, + -1, + 0, + 2048 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_200", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_90" + }, + -1, + 2048, + 4096 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_201", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_90" + }, + -1, + 4096, + 6144 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 486, in forward\n query, key, value = torch.split(\n ^^^^^^^^^^^^" + }, + { + "id": "view_145", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_199" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 496, in forward\n query = query.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_146", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_200" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 497, in forward\n key = key.reshape(batch_size, seq_len, -1, self.head_k_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_147", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_201" + }, + [ + 1, + 64, + 16, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 498, in forward\n value = value.reshape(batch_size, seq_len, -1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_22", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "linear_173" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 500, in forward\n beta = b.sigmoid()\n ^^^^^^^^^^^" + }, + { + "id": "exp_17", + "op": "aten.exp.default", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_a_log" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_27", + "op": "aten.neg.default", + "args": [ + { + "tensor": "exp_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_237", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "linear_174" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_dt_bias" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softplus_17", + "op": "aten.softplus.default", + "args": [ + { + "tensor": "add_237" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_300", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "neg_27" + }, + { + "tensor": "softplus_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 502, in forward\n g = -self.A_log.float().exp() * F.softplus(a.float() + self.dt_bias)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_301", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_145" + }, + { + "tensor": "view_145" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_35", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_301" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_238", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_35" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_106", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_238" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_302", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_145" + }, + { + "tensor": "rsqrt_106" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 26, in _webtorch_chunk_gated_delta_rule\n query = l2norm(query, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "mul_303", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_146" + }, + { + "tensor": "view_146" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sum_36", + "op": "aten.sum.dim_IntList", + "args": [ + { + "tensor": "mul_303" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_239", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "sum_36" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_107", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_239" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 244, in l2norm\n inv_norm = torch.rsqrt((x * x).sum(dim=dim, keepdim=True) + eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_304", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_146" + }, + { + "tensor": "rsqrt_107" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 27, in _webtorch_chunk_gated_delta_rule\n key = l2norm(key, dim=-1, eps=1e-6)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 245, in l2norm\n return x * inv_norm\n ^^^^^^^^^^^^" + }, + { + "id": "normalized_gated_delta_rule_17", + "op": "webtorch.normalized_gated_delta_rule.default", + "args": [ + { + "tensor": "mul_302" + }, + { + "tensor": "mul_304" + }, + { + "tensor": "view_147" + }, + { + "tensor": "mul_300" + }, + { + "tensor": "sigmoid_22" + }, + { + "tensor": "states_45" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 28, in _webtorch_chunk_gated_delta_rule\n packed = normalized_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_202", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_17" + }, + 2, + 0, + 128 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 32, in _webtorch_chunk_gated_delta_rule\n final_state = packed[:, :, :key_width]\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_203", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "normalized_gated_delta_rule_17" + }, + 2, + 128, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_91", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "slice_203" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_37", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_91" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 522, in forward\n core_attn_out, last_recurrent_state = torch_chunk_gated_delta_rule(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 33, in _webtorch_chunk_gated_delta_rule\n output = packed[:, :, key_width:].transpose(1, 2).contiguous().to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_148", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_37" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 540, in forward\n core_attn_out = core_attn_out.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_149", + "op": "aten.view.default", + "args": [ + { + "tensor": "view_144" + }, + [ + -1, + 128 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 541, in forward\n z = z.reshape(-1, self.head_v_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_73", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_148" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_72", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_73" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 178, in forward\n variance = hidden_states.pow(2).mean(-1, keepdim=True)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_240", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_72" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_108", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_240" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_305", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_148" + }, + { + "tensor": "rsqrt_108" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 180, in forward\n hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_306", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_norm_weight" + }, + { + "tensor": "mul_305" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 181, in forward\n hidden_states = self.weight * hidden_states.to(input_dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_57", + "op": "aten.silu.default", + "args": [ + { + "tensor": "view_149" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_307", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_306" + }, + { + "tensor": "silu_57" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 542, in forward\n core_attn_out = self.norm(core_attn_out, z)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 182, in forward\n hidden_states = hidden_states * ACT2FN[self.activation](gate.to(torch.float32))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_150", + "op": "aten.view.default", + "args": [ + { + "tensor": "mul_307" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 543, in forward\n core_attn_out = core_attn_out.reshape(batch_size, seq_len, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_175", + "op": "aten.linear.default", + "args": [ + { + "tensor": "view_150" + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_out_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 771, in forward\n hidden_states = self.linear_attn(\n ^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/integrations/accelerate.py\", line 1010, in wrapped\n output = forward_func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 545, in forward\n output = self.out_proj(core_attn_out)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_241", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_234" + }, + { + "tensor": "linear_175" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_74", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_241" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_73", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_74" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_242", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_73" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_109", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_242" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_308", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_241" + }, + { + "tensor": "rsqrt_109" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_243", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_22_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_309", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_308" + }, + { + "tensor": "add_243" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_55", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_309" + }, + { + "tensor": "add_241" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_176", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_55" + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_58", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_176" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_177", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_55" + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_310", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_58" + }, + { + "tensor": "linear_177" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_178", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_310" + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_244", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_241" + }, + { + "tensor": "linear_178" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_75", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_244" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_74", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_75" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_245", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_74" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_110", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_245" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_311", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_244" + }, + { + "tensor": "rsqrt_110" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_246", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_input_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_312", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_311" + }, + { + "tensor": "add_246" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_56", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_312" + }, + { + "tensor": "add_244" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 767, in forward\n hidden_states = self.input_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_179", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_56" + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_q_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "view_151", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_179" + }, + [ + 1, + 64, + -1, + 512 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 671, in forward\n self.q_proj(hidden_states).view(*input_shape, -1, self.head_dim * 2), 2, dim=-1" + }, + { + "id": "slice_204", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_151" + }, + -1, + 0, + 256 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "slice_205", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "view_151" + }, + -1, + 256, + 512 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 670, in forward\n query_states, gate = torch.chunk(\n ^^^^^^^^^^^^" + }, + { + "id": "clone_38", + "op": "aten.clone.default", + "args": [ + { + "tensor": "slice_205" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_15", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_38" + }, + [ + 1, + 64, + 2048 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 673, in forward\n gate = gate.reshape(*input_shape, -1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_152", + "op": "aten.view.default", + "args": [ + { + "tensor": "slice_204" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_76", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_152" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_75", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_76" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_247", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_75" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_111", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_247" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_313", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_152" + }, + { + "tensor": "rsqrt_111" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_248", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_self_attn_q_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_314", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_313" + }, + { + "tensor": "add_248" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_57", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_314" + }, + { + "tensor": "view_152" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_92", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_57" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 675, in forward\n query_states = self.q_norm(query_states.view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_180", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_56" + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_k_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_153", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_180" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_77", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "view_153" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_76", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_77" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_249", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_76" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_112", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_249" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_315", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_153" + }, + { + "tensor": "rsqrt_112" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_250", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_self_attn_k_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_316", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_315" + }, + { + "tensor": "add_250" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_58", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_316" + }, + { + "tensor": "view_153" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_93", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "type_as_58" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 676, in forward\n key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_181", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_56" + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_v_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_154", + "op": "aten.view.default", + "args": [ + { + "tensor": "linear_181" + }, + [ + 1, + 64, + -1, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_94", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "view_154" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 677, in forward\n value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_43", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 577, in apply_rotary_pos_emb\n cos = cos.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_44", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "mul_1" + }, + 1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 578, in apply_rotary_pos_emb\n sin = sin.unsqueeze(unsqueeze_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_206", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_92" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_207", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_92" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 582, in apply_rotary_pos_emb\n q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_208", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_93" + }, + 3, + 0, + 64 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_209", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "transpose_93" + }, + 3, + 64, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 583, in apply_rotary_pos_emb\n k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_317", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_206" + }, + { + "tensor": "unsqueeze_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_210", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_206" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_211", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_206" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_28", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_211" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_39", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_28" + }, + { + "tensor": "slice_210" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_318", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_39" + }, + { + "tensor": "unsqueeze_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_251", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_317" + }, + { + "tensor": "mul_318" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 586, in apply_rotary_pos_emb\n q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_319", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "slice_208" + }, + { + "tensor": "unsqueeze_43" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_212", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_208" + }, + 3, + 0, + 32 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 551, in rotate_half\n x1 = x[..., : x.shape[-1] // 2]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "slice_213", + "op": "aten.slice.Tensor", + "args": [ + { + "tensor": "slice_208" + }, + 3, + 32, + 9223372036854775807 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 552, in rotate_half\n x2 = x[..., x.shape[-1] // 2 :]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "neg_29", + "op": "aten.neg.default", + "args": [ + { + "tensor": "slice_213" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_40", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "neg_29" + }, + { + "tensor": "slice_212" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 553, in rotate_half\n return torch.cat((-x2, x1), dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_320", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "cat_40" + }, + { + "tensor": "unsqueeze_44" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_252", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_319" + }, + { + "tensor": "mul_320" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 587, in apply_rotary_pos_emb\n k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_41", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_251" + }, + { + "tensor": "slice_207" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 590, in apply_rotary_pos_emb\n q_embed = torch.cat([q_embed, q_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cat_42", + "op": "aten.cat.default", + "args": [ + [ + { + "tensor": "add_252" + }, + { + "tensor": "slice_209" + } + ], + -1 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 680, in forward\n query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 591, in apply_rotary_pos_emb\n k_embed = torch.cat([k_embed, k_pass], dim=-1)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_10", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_46" + }, + { + "tensor": "cat_42" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 91, in update\n self.states[2 * layer_idx] = cache_update(self.states[2 * layer_idx], key, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "cache_update_11", + "op": "webtorch.cache_update.default", + "args": [ + { + "tensor": "states_47" + }, + { + "tensor": "transpose_94" + }, + { + "tensor": "position_ids" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 683, in forward\n key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 92, in update\n self.states[2 * layer_idx + 1] = cache_update(self.states[2 * layer_idx + 1], value, self.positions)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_45", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_10" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_12", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_45" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_39", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_12" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_16", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_39" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 617, in eager_attention_forward\n key_states = repeat_kv(key, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "unsqueeze_46", + "op": "aten.unsqueeze.default", + "args": [ + { + "tensor": "cache_update_11" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "expand_13", + "op": "aten.expand.default", + "args": [ + { + "tensor": "unsqueeze_46" + }, + [ + 1, + 2, + 4, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 603, in repeat_kv\n hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_40", + "op": "aten.clone.default", + "args": [ + { + "tensor": "expand_13" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "_unsafe_view_17", + "op": "aten._unsafe_view.default", + "args": [ + { + "tensor": "clone_40" + }, + [ + 1, + 8, + 4096, + 256 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 618, in eager_attention_forward\n value_states = repeat_kv(value, module.num_key_value_groups)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 604, in repeat_kv\n return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_95", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "_unsafe_view_16" + }, + 2, + 3 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_11", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "cat_41" + }, + { + "tensor": "transpose_95" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_321", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "matmul_11" + }, + 0.0625 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 620, in eager_attention_forward\n attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_253", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mul_321" + }, + { + "tensor": "attention_mask" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 622, in eager_attention_forward\n attn_weights = attn_weights + attention_mask\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "softmax_5", + "op": "aten.softmax.int", + "args": [ + { + "tensor": "add_253" + }, + -1, + { + "dtype": "float32" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 624, in eager_attention_forward\n attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "matmul_12", + "op": "aten.matmul.default", + "args": [ + { + "tensor": "softmax_5" + }, + { + "tensor": "_unsafe_view_17" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 626, in eager_attention_forward\n attn_output = torch.matmul(attn_weights, value_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "transpose_96", + "op": "aten.transpose.int", + "args": [ + { + "tensor": "matmul_12" + }, + 1, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "clone_41", + "op": "aten.clone.default", + "args": [ + { + "tensor": "transpose_96" + } + ], + "kwargs": { + "memory_format": { + "memoryFormat": "contiguous_format" + } + }, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 689, in forward\n attn_output, attn_weights = attention_interface(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 627, in eager_attention_forward\n attn_output = attn_output.transpose(1, 2).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "view_155", + "op": "aten.view.default", + "args": [ + { + "tensor": "clone_41" + }, + [ + 1, + 64, + -1 + ] + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 700, in forward\n attn_output = attn_output.reshape(*input_shape, -1).contiguous()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "sigmoid_23", + "op": "aten.sigmoid.default", + "args": [ + { + "tensor": "_unsafe_view_15" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_322", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "view_155" + }, + { + "tensor": "sigmoid_23" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 701, in forward\n attn_output = attn_output * torch.sigmoid(gate)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_182", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_322" + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_o_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 779, in forward\n hidden_states, _ = self.self_attn(\n ^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 703, in forward\n attn_output = self.o_proj(attn_output)\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_254", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_244" + }, + { + "tensor": "linear_182" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 788, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_78", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_254" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_77", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_78" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_255", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_77" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_113", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_255" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_323", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_254" + }, + { + "tensor": "rsqrt_113" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_256", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_layers_23_post_attention_layernorm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_324", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_323" + }, + { + "tensor": "add_256" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_59", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_324" + }, + { + "tensor": "add_254" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 792, in forward\n hidden_states = self.post_attention_layernorm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_183", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_59" + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_gate_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "silu_59", + "op": "aten.silu.default", + "args": [ + { + "tensor": "linear_183" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/activations.py\", line 103, in forward\n return nn.functional.silu(input)\n ^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_184", + "op": "aten.linear.default", + "args": [ + { + "tensor": "type_as_59" + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_up_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_325", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "silu_59" + }, + { + "tensor": "linear_184" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_185", + "op": "aten.linear.default", + "args": [ + { + "tensor": "mul_325" + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_down_proj_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 793, in forward\n hidden_states = self.mlp(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 719, in forward\n down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_257", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "add_254" + }, + { + "tensor": "linear_185" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1206, in forward\n hidden_states = decoder_layer(\n ^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/modeling_layers.py\", line 114, in __call__\n return super().__call__(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 794, in forward\n hidden_states = residual + hidden_states\n ^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "pow_79", + "op": "aten.pow.Tensor_Scalar", + "args": [ + { + "tensor": "add_257" + }, + 2 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mean_78", + "op": "aten.mean.dim", + "args": [ + { + "tensor": "pow_79" + }, + [ + -1 + ], + true + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_258", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "mean_78" + }, + 1e-06 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "rsqrt_114", + "op": "aten.rsqrt.default", + "args": [ + { + "tensor": "add_258" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_326", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "add_257" + }, + { + "tensor": "rsqrt_114" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 733, in forward\n output = self._norm(x.float())\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 730, in _norm\n return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "add_259", + "op": "aten.add.Tensor", + "args": [ + { + "tensor": "p_model_model_language_model_norm_weight" + }, + 1.0 + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "mul_327", + "op": "aten.mul.Tensor", + "args": [ + { + "tensor": "mul_326" + }, + { + "tensor": "add_259" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 736, in forward\n output = output * (1.0 + self.weight.float())\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "type_as_60", + "op": "aten.type_as.default", + "args": [ + { + "tensor": "mul_327" + }, + { + "tensor": "add_257" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1769, in forward\n outputs = self.model(\n ^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1562, in forward\n outputs = self.language_model(\n ^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 1068, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/output_capturing.py\", line 262, in wrapper\n outputs = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1216, in forward\n hidden_states = self.norm(hidden_states)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 737, in forward\n return output.type_as(x)\n ^^^^^^^^^^^^^^^^^" + }, + { + "id": "alias", + "op": "aten.alias.default", + "args": [ + { + "tensor": "type_as_60" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1787, in forward\n logits = self.lm_head(hidden_states[:, slice_indices, :])\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + }, + { + "id": "linear_186", + "op": "aten.linear.default", + "args": [ + { + "tensor": "alias" + }, + { + "tensor": "p_model_model_language_model_embed_tokens_weight" + } + ], + "kwargs": {}, + "source": "File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/torch/_dynamo/functional_export.py\", line 243, in forward\n res = self._export_root(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 135, in forward\n return self._forward(input_ids, position_ids, attention_mask, states)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/python/webtorch/adapters/qwen35.py\", line 114, in _forward\n output = self.model(input_ids=input_ids, position_ids=position_ids,\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/utils/generic.py\", line 939, in wrapper\n output = func(self, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/Users/reserve/Workspace/webtorch/.venv/lib/python3.11/site-packages/transformers/models/qwen3_5/modeling_qwen3_5.py\", line 1787, in forward\n logits = self.lm_head(hidden_states[:, slice_indices, :])\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" + } + ], + "quantization": [ + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007083918899297714, + "meanSquaredWeightError": 4.075269544046023e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005190409719944, + "meanSquaredWeightError": 1.0724062526890066e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014917552471160889, + "meanSquaredWeightError": 8.557733188752081e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001733013428747654, + "meanSquaredWeightError": 9.211684748322568e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_0_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001013568602502346, + "meanSquaredWeightError": 4.7989616724919415e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0420619398355484, + "meanSquaredWeightError": 9.975725561162108e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03571777045726776, + "meanSquaredWeightError": 1.8712555629463168e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_0_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.015869140625, + "meanSquaredWeightError": 9.336113180324901e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006478391587734222, + "meanSquaredWeightError": 1.841971197791281e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006430316716432571, + "meanSquaredWeightError": 1.099748647703791e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009914608672261238, + "meanSquaredWeightError": 8.36655100755479e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006439946591854095, + "meanSquaredWeightError": 5.678034487033301e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_10_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010154899209737778, + "meanSquaredWeightError": 4.161362809895763e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04477539658546448, + "meanSquaredWeightError": 9.298059922002722e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02817993052303791, + "meanSquaredWeightError": 1.585794393577089e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_10_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.017333984375, + "meanSquaredWeightError": 9.087384000849852e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03647809475660324, + "meanSquaredWeightError": 9.2817742824991e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03266602009534836, + "meanSquaredWeightError": 1.564963326927682e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_11_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0213623046875, + "meanSquaredWeightError": 9.297695555687824e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00043938105227425694, + "meanSquaredWeightError": 4.2963801405448976e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0004978924989700317, + "meanSquaredWeightError": 4.300909850485368e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007458776235580444, + "meanSquaredWeightError": 6.14908657325941e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_11_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005382634699344635, + "meanSquaredWeightError": 6.678897435108411e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006161201745271683, + "meanSquaredWeightError": 2.1588792975535398e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0004498325288295746, + "meanSquaredWeightError": 7.733858886638245e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010226964950561523, + "meanSquaredWeightError": 1.0263062399928913e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007535666227340698, + "meanSquaredWeightError": 6.758249515570469e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_12_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007804818451404572, + "meanSquaredWeightError": 3.901024836494571e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02197265625, + "meanSquaredWeightError": 8.842465035741043e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03532366454601288, + "meanSquaredWeightError": 1.9045186263610958e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_12_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01679687388241291, + "meanSquaredWeightError": 8.847858339322556e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006036236882209778, + "meanSquaredWeightError": 1.5833309774393456e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00047482550144195557, + "meanSquaredWeightError": 7.293367687566388e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009035095572471619, + "meanSquaredWeightError": 9.150482149777872e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007492424920201302, + "meanSquaredWeightError": 6.225117310521e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_13_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0012094092089682817, + "meanSquaredWeightError": 4.19806900353592e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0224609375, + "meanSquaredWeightError": 9.161611842500861e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04632568359375, + "meanSquaredWeightError": 2.056171069853008e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_13_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0159912109375, + "meanSquaredWeightError": 9.076159699361597e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007881708443164825, + "meanSquaredWeightError": 1.647453373720964e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005805529654026031, + "meanSquaredWeightError": 7.469957985506426e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0016993731260299683, + "meanSquaredWeightError": 9.388108956898122e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007583741098642349, + "meanSquaredWeightError": 7.161021109425292e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_14_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006665807450190187, + "meanSquaredWeightError": 4.480360082936841e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.037353515625, + "meanSquaredWeightError": 1.04351170193695e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.05363769829273224, + "meanSquaredWeightError": 2.7039927772420924e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_14_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01563110202550888, + "meanSquaredWeightError": 9.754314760357374e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.047607421875, + "meanSquaredWeightError": 1.168099856840854e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.05196359008550644, + "meanSquaredWeightError": 2.913413027272327e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_15_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0177001953125, + "meanSquaredWeightError": 1.0753815331554506e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00047194119542837143, + "meanSquaredWeightError": 4.19777368421137e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006257314234972, + "meanSquaredWeightError": 6.4926997112024765e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000988095998764038, + "meanSquaredWeightError": 7.445749794499079e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_15_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0011749267578125, + "meanSquaredWeightError": 1.0600491151535607e-08, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005680583417415619, + "meanSquaredWeightError": 1.9959742303399253e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005613304674625397, + "meanSquaredWeightError": 9.642544540611198e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0013072118163108826, + "meanSquaredWeightError": 1.371929148064055e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000817006453871727, + "meanSquaredWeightError": 7.562183768072828e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_16_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010419189929962158, + "meanSquaredWeightError": 4.178038359725633e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0301513671875, + "meanSquaredWeightError": 1.0534593002375914e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03706229105591774, + "meanSquaredWeightError": 3.2628909139020834e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_16_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01273193396627903, + "meanSquaredWeightError": 1.0341716460970929e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008150823414325714, + "meanSquaredWeightError": 2.1580238040996846e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00043829530477523804, + "meanSquaredWeightError": 7.784422884071773e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014495849609375, + "meanSquaredWeightError": 1.5511064432871535e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000626690685749054, + "meanSquaredWeightError": 8.258425054918916e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_17_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006689839065074921, + "meanSquaredWeightError": 4.077417070647016e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0342668816447258, + "meanSquaredWeightError": 1.0280851938659907e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.046142578125, + "meanSquaredWeightError": 3.7489039641513955e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_17_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0156511589884758, + "meanSquaredWeightError": 1.021842535919859e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007631815969944, + "meanSquaredWeightError": 1.8714940708264294e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005142334848642349, + "meanSquaredWeightError": 8.547659469115843e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014417767524719238, + "meanSquaredWeightError": 1.6978646044663037e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000848725438117981, + "meanSquaredWeightError": 8.861574585239396e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_18_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009453240782022476, + "meanSquaredWeightError": 3.604574860816001e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0531267449259758, + "meanSquaredWeightError": 1.0224184734397568e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04545222967863083, + "meanSquaredWeightError": 4.110375357413432e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_18_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.014916989952325821, + "meanSquaredWeightError": 1.0372635870226077e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04470040649175644, + "meanSquaredWeightError": 1.1175421832376742e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04736328125, + "meanSquaredWeightError": 3.7854479160159826e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_19_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01680908165872097, + "meanSquaredWeightError": 1.0865893500522361e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006805174052715302, + "meanSquaredWeightError": 4.193761338200375e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008737146854400635, + "meanSquaredWeightError": 9.042077309118213e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007929764688014984, + "meanSquaredWeightError": 7.600540641305997e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_19_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014283191412687302, + "meanSquaredWeightError": 1.78832415542729e-08, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000786246731877327, + "meanSquaredWeightError": 2.793629150232846e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008093155920505524, + "meanSquaredWeightError": 1.2136558424913346e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0012167382519692183, + "meanSquaredWeightError": 8.249294580764399e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007112771272659302, + "meanSquaredWeightError": 1.0718528287156914e-08, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_1_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0012726113200187683, + "meanSquaredWeightError": 4.84854734139617e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04365234822034836, + "meanSquaredWeightError": 9.266742040381359e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.05205077677965164, + "meanSquaredWeightError": 2.190386794609367e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_1_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02226562425494194, + "meanSquaredWeightError": 9.115561283579154e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007804818451404572, + "meanSquaredWeightError": 2.823620093295176e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0004916451871395111, + "meanSquaredWeightError": 1.1404850397411792e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0022107213735580444, + "meanSquaredWeightError": 1.7722628697924847e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001212533563375473, + "meanSquaredWeightError": 9.309404802593235e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_20_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001015007495880127, + "meanSquaredWeightError": 3.86098308879923e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03144530951976776, + "meanSquaredWeightError": 1.1349436590535333e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.043701171875, + "meanSquaredWeightError": 3.7038337268313626e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_20_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01273193396627903, + "meanSquaredWeightError": 1.1466236173873767e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000811239704489708, + "meanSquaredWeightError": 2.8647622940525252e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0004772273823618889, + "meanSquaredWeightError": 9.231112763075089e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001953125, + "meanSquaredWeightError": 1.5947463793963834e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010938271880149841, + "meanSquaredWeightError": 9.523899890950815e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_21_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010099634528160095, + "meanSquaredWeightError": 4.4895744899520196e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02999267354607582, + "meanSquaredWeightError": 1.2448520010366337e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03925780951976776, + "meanSquaredWeightError": 3.853431735478807e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_21_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02015380933880806, + "meanSquaredWeightError": 1.2184170827822527e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008477643132209778, + "meanSquaredWeightError": 3.4146133032209036e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007651001214981079, + "meanSquaredWeightError": 1.287605755351251e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0015244372189044952, + "meanSquaredWeightError": 1.2338504440378983e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001476377248764038, + "meanSquaredWeightError": 9.998956329582143e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_22_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001247616484761238, + "meanSquaredWeightError": 5.051828733115826e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0316162109375, + "meanSquaredWeightError": 1.3124330280334107e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.05973249301314354, + "meanSquaredWeightError": 4.17917181039229e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_22_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02794799953699112, + "meanSquaredWeightError": 1.3120700259605655e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0629359632730484, + "meanSquaredWeightError": 1.335452111561608e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0677926167845726, + "meanSquaredWeightError": 4.988436103303684e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_23_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0322265625, + "meanSquaredWeightError": 1.6804390270408476e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000582478940486908, + "meanSquaredWeightError": 3.9153045250373e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010428838431835175, + "meanSquaredWeightError": 8.011205920865905e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001420147716999054, + "meanSquaredWeightError": 8.69789218427286e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_23_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.002052607014775276, + "meanSquaredWeightError": 1.3341496796215324e-08, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0007699094712734222, + "meanSquaredWeightError": 2.5097170563981308e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006843628361821175, + "meanSquaredWeightError": 1.136356786446413e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009746402502059937, + "meanSquaredWeightError": 8.813485941061572e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000755971297621727, + "meanSquaredWeightError": 1.1121874976538493e-08, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_2_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0011303499341011047, + "meanSquaredWeightError": 4.848550450020639e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03278808295726776, + "meanSquaredWeightError": 8.690146273693244e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0361328125, + "meanSquaredWeightError": 3.182140517310472e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_2_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01421508751809597, + "meanSquaredWeightError": 8.130265882755339e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03572387620806694, + "meanSquaredWeightError": 8.140267482303898e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02766112983226776, + "meanSquaredWeightError": 3.193840257154079e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_3_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01599992997944355, + "meanSquaredWeightError": 7.776296229167201e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00040658097714185715, + "meanSquaredWeightError": 4.779415085920391e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0012706853449344635, + "meanSquaredWeightError": 4.892196425743123e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009188912808895111, + "meanSquaredWeightError": 1.1496103624608622e-08, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_3_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005174772813916206, + "meanSquaredWeightError": 5.1614277296607725e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006728284060955048, + "meanSquaredWeightError": 2.26705694217344e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005107489414513111, + "meanSquaredWeightError": 8.198658640878875e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001364879310131073, + "meanSquaredWeightError": 1.0674462203041912e-08, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006516836583614349, + "meanSquaredWeightError": 9.449681925843834e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_4_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009464053437113762, + "meanSquaredWeightError": 4.404944853320103e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0265415720641613, + "meanSquaredWeightError": 8.238882287514571e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0333251990377903, + "meanSquaredWeightError": 2.927112973338808e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_4_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0194091796875, + "meanSquaredWeightError": 8.087501441877976e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005795937031507492, + "meanSquaredWeightError": 1.80652719450336e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005795937031507492, + "meanSquaredWeightError": 7.52482787191866e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0013129785656929016, + "meanSquaredWeightError": 9.38157196372913e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008208509534597397, + "meanSquaredWeightError": 8.128849593447285e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_5_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010572969913482666, + "meanSquaredWeightError": 4.412351817251192e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04150390625, + "meanSquaredWeightError": 8.620316975793685e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0302734375, + "meanSquaredWeightError": 2.6132290713576367e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_5_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02303466945886612, + "meanSquaredWeightError": 8.573984473514429e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0006062665488570929, + "meanSquaredWeightError": 1.623409140449894e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005637351423501968, + "meanSquaredWeightError": 9.469766304448513e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008612200617790222, + "meanSquaredWeightError": 8.620569147410606e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008112378418445587, + "meanSquaredWeightError": 7.537758861531074e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_6_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008592978119850159, + "meanSquaredWeightError": 4.527719088542881e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.028881840407848358, + "meanSquaredWeightError": 9.06856655547017e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.04156494140625, + "meanSquaredWeightError": 2.46402237280563e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_6_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.01636962778866291, + "meanSquaredWeightError": 9.115400416703778e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02463379129767418, + "meanSquaredWeightError": 8.773882314017101e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.030029296875, + "meanSquaredWeightError": 1.9424107904342236e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_7_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02099609375, + "meanSquaredWeightError": 9.083250915864483e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_k_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00036885414738208055, + "meanSquaredWeightError": 3.3473028793196136e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_o_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005767084658145905, + "meanSquaredWeightError": 3.908651180495326e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_q_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 4194304, + "groupSize": 32, + "maxAbsoluteWeightError": 0.001015007495880127, + "meanSquaredWeightError": 7.261003798220145e-09, + "storageBytes": 4718592 + }, + { + "tensor": "p_model_model_language_model_layers_7_self_attn_v_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 524288, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005060629919171333, + "meanSquaredWeightError": 4.816435694721122e-09, + "storageBytes": 589824 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005863220430910587, + "meanSquaredWeightError": 1.6181449069563314e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.00045013427734375, + "meanSquaredWeightError": 6.057955026506079e-09, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009198524057865143, + "meanSquaredWeightError": 8.711624310819843e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0013682451099157333, + "meanSquaredWeightError": 6.523535933666835e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_8_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0008987076580524445, + "meanSquaredWeightError": 4.059202307615806e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.037841796875, + "meanSquaredWeightError": 8.265481596936297e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.03264160454273224, + "meanSquaredWeightError": 1.6808718328320538e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_8_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0196533203125, + "meanSquaredWeightError": 8.899761496650171e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.000939076766371727, + "meanSquaredWeightError": 2.0765920538678984e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 16384, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0005536414682865143, + "meanSquaredWeightError": 1.1359857943205043e-08, + "storageBytes": 18432 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 6291456, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0010246187448501587, + "meanSquaredWeightError": 7.989317651890815e-09, + "storageBytes": 7077888 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0014816643670201302, + "meanSquaredWeightError": 6.387295137244564e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_9_linear_attn_out_proj_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 2097152, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009112022817134857, + "meanSquaredWeightError": 3.856152730463691e-09, + "storageBytes": 2359296 + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_down_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0225830078125, + "meanSquaredWeightError": 8.832107027956226e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_gate_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.02226562611758709, + "meanSquaredWeightError": 1.4517542012981721e-06, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_layers_9_mlp_up_proj_weight", + "scheme": "symmetric-int4", + "method": "gptq", + "elements": 3670016, + "groupSize": 32, + "maxAbsoluteWeightError": 0.017578125, + "meanSquaredWeightError": 9.189842558043892e-07, + "storageBytes": 2293760 + }, + { + "tensor": "p_model_model_language_model_embed_tokens_weight", + "scheme": "symmetric-int8", + "method": "rtn", + "elements": 254279680, + "groupSize": 32, + "maxAbsoluteWeightError": 0.0009068762883543968, + "meanSquaredWeightError": 1.1976288405435298e-08, + "storageBytes": 286064640 + } + ], + "quantizationTiers": [ + { + "scheme": "symmetric-int8", + "method": "rtn", + "groupSize": 32, + "include": [], + "exclude": [], + "matrices": 0 + }, + { + "scheme": "symmetric-int4", + "method": "gptq", + "groupSize": 32, + "include": [ + "p_model_model_language_model_layers_0_mlp_down_proj_weight", + "p_model_model_language_model_layers_0_mlp_gate_proj_weight", + "p_model_model_language_model_layers_0_mlp_up_proj_weight", + "p_model_model_language_model_layers_10_mlp_down_proj_weight", + "p_model_model_language_model_layers_10_mlp_gate_proj_weight", + "p_model_model_language_model_layers_10_mlp_up_proj_weight", + "p_model_model_language_model_layers_11_mlp_down_proj_weight", + "p_model_model_language_model_layers_11_mlp_gate_proj_weight", + "p_model_model_language_model_layers_11_mlp_up_proj_weight", + "p_model_model_language_model_layers_12_mlp_down_proj_weight", + "p_model_model_language_model_layers_12_mlp_gate_proj_weight", + "p_model_model_language_model_layers_12_mlp_up_proj_weight", + "p_model_model_language_model_layers_13_mlp_down_proj_weight", + "p_model_model_language_model_layers_13_mlp_gate_proj_weight", + "p_model_model_language_model_layers_13_mlp_up_proj_weight", + "p_model_model_language_model_layers_14_mlp_down_proj_weight", + "p_model_model_language_model_layers_14_mlp_gate_proj_weight", + "p_model_model_language_model_layers_14_mlp_up_proj_weight", + "p_model_model_language_model_layers_15_mlp_down_proj_weight", + "p_model_model_language_model_layers_15_mlp_gate_proj_weight", + "p_model_model_language_model_layers_15_mlp_up_proj_weight", + "p_model_model_language_model_layers_16_mlp_down_proj_weight", + "p_model_model_language_model_layers_16_mlp_gate_proj_weight", + "p_model_model_language_model_layers_16_mlp_up_proj_weight", + "p_model_model_language_model_layers_17_mlp_down_proj_weight", + "p_model_model_language_model_layers_17_mlp_gate_proj_weight", + "p_model_model_language_model_layers_17_mlp_up_proj_weight", + "p_model_model_language_model_layers_18_mlp_down_proj_weight", + "p_model_model_language_model_layers_18_mlp_gate_proj_weight", + "p_model_model_language_model_layers_18_mlp_up_proj_weight", + "p_model_model_language_model_layers_19_mlp_down_proj_weight", + "p_model_model_language_model_layers_19_mlp_gate_proj_weight", + "p_model_model_language_model_layers_19_mlp_up_proj_weight", + "p_model_model_language_model_layers_1_mlp_down_proj_weight", + "p_model_model_language_model_layers_1_mlp_gate_proj_weight", + "p_model_model_language_model_layers_1_mlp_up_proj_weight", + "p_model_model_language_model_layers_20_mlp_down_proj_weight", + "p_model_model_language_model_layers_20_mlp_gate_proj_weight", + "p_model_model_language_model_layers_20_mlp_up_proj_weight", + "p_model_model_language_model_layers_21_mlp_down_proj_weight", + "p_model_model_language_model_layers_21_mlp_gate_proj_weight", + "p_model_model_language_model_layers_21_mlp_up_proj_weight", + "p_model_model_language_model_layers_22_mlp_down_proj_weight", + "p_model_model_language_model_layers_22_mlp_gate_proj_weight", + "p_model_model_language_model_layers_22_mlp_up_proj_weight", + "p_model_model_language_model_layers_23_mlp_down_proj_weight", + "p_model_model_language_model_layers_23_mlp_gate_proj_weight", + "p_model_model_language_model_layers_23_mlp_up_proj_weight", + "p_model_model_language_model_layers_2_mlp_down_proj_weight", + "p_model_model_language_model_layers_2_mlp_gate_proj_weight", + "p_model_model_language_model_layers_2_mlp_up_proj_weight", + "p_model_model_language_model_layers_3_mlp_down_proj_weight", + "p_model_model_language_model_layers_3_mlp_gate_proj_weight", + "p_model_model_language_model_layers_3_mlp_up_proj_weight", + "p_model_model_language_model_layers_4_mlp_down_proj_weight", + "p_model_model_language_model_layers_4_mlp_gate_proj_weight", + "p_model_model_language_model_layers_4_mlp_up_proj_weight", + "p_model_model_language_model_layers_5_mlp_down_proj_weight", + "p_model_model_language_model_layers_5_mlp_gate_proj_weight", + "p_model_model_language_model_layers_5_mlp_up_proj_weight", + "p_model_model_language_model_layers_6_mlp_down_proj_weight", + "p_model_model_language_model_layers_6_mlp_gate_proj_weight", + "p_model_model_language_model_layers_6_mlp_up_proj_weight", + "p_model_model_language_model_layers_7_mlp_down_proj_weight", + "p_model_model_language_model_layers_7_mlp_gate_proj_weight", + "p_model_model_language_model_layers_7_mlp_up_proj_weight", + "p_model_model_language_model_layers_8_mlp_down_proj_weight", + "p_model_model_language_model_layers_8_mlp_gate_proj_weight", + "p_model_model_language_model_layers_8_mlp_up_proj_weight", + "p_model_model_language_model_layers_9_mlp_down_proj_weight", + "p_model_model_language_model_layers_9_mlp_gate_proj_weight", + "p_model_model_language_model_layers_9_mlp_up_proj_weight" + ], + "exclude": [], + "matrices": 72 + }, + { + "scheme": "symmetric-int8", + "method": "rtn", + "groupSize": 32, + "include": [ + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_11_self_attn_k_proj_weight", + "p_model_model_language_model_layers_11_self_attn_o_proj_weight", + "p_model_model_language_model_layers_11_self_attn_q_proj_weight", + "p_model_model_language_model_layers_11_self_attn_v_proj_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_15_self_attn_k_proj_weight", + "p_model_model_language_model_layers_15_self_attn_o_proj_weight", + "p_model_model_language_model_layers_15_self_attn_q_proj_weight", + "p_model_model_language_model_layers_15_self_attn_v_proj_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_19_self_attn_k_proj_weight", + "p_model_model_language_model_layers_19_self_attn_o_proj_weight", + "p_model_model_language_model_layers_19_self_attn_q_proj_weight", + "p_model_model_language_model_layers_19_self_attn_v_proj_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_23_self_attn_k_proj_weight", + "p_model_model_language_model_layers_23_self_attn_o_proj_weight", + "p_model_model_language_model_layers_23_self_attn_q_proj_weight", + "p_model_model_language_model_layers_23_self_attn_v_proj_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_3_self_attn_k_proj_weight", + "p_model_model_language_model_layers_3_self_attn_o_proj_weight", + "p_model_model_language_model_layers_3_self_attn_q_proj_weight", + "p_model_model_language_model_layers_3_self_attn_v_proj_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_7_self_attn_k_proj_weight", + "p_model_model_language_model_layers_7_self_attn_o_proj_weight", + "p_model_model_language_model_layers_7_self_attn_q_proj_weight", + "p_model_model_language_model_layers_7_self_attn_v_proj_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight" + ], + "exclude": [], + "matrices": 114 + } + ], + "states": [ + { + "input": "states_0", + "output": "slice_10", + "kind": "convolution" + }, + { + "input": "states_1", + "output": "slice_16", + "kind": "recurrent" + }, + { + "input": "states_2", + "output": "slice_18", + "kind": "convolution" + }, + { + "input": "states_3", + "output": "slice_24", + "kind": "recurrent" + }, + { + "input": "states_4", + "output": "slice_26", + "kind": "convolution" + }, + { + "input": "states_5", + "output": "slice_32", + "kind": "recurrent" + }, + { + "input": "states_6", + "output": "cache_update", + "kind": "kv" + }, + { + "input": "states_7", + "output": "cache_update_1", + "kind": "kv" + }, + { + "input": "states_8", + "output": "slice_44", + "kind": "convolution" + }, + { + "input": "states_9", + "output": "slice_50", + "kind": "recurrent" + }, + { + "input": "states_10", + "output": "slice_52", + "kind": "convolution" + }, + { + "input": "states_11", + "output": "slice_58", + "kind": "recurrent" + }, + { + "input": "states_12", + "output": "slice_60", + "kind": "convolution" + }, + { + "input": "states_13", + "output": "slice_66", + "kind": "recurrent" + }, + { + "input": "states_14", + "output": "cache_update_2", + "kind": "kv" + }, + { + "input": "states_15", + "output": "cache_update_3", + "kind": "kv" + }, + { + "input": "states_16", + "output": "slice_78", + "kind": "convolution" + }, + { + "input": "states_17", + "output": "slice_84", + "kind": "recurrent" + }, + { + "input": "states_18", + "output": "slice_86", + "kind": "convolution" + }, + { + "input": "states_19", + "output": "slice_92", + "kind": "recurrent" + }, + { + "input": "states_20", + "output": "slice_94", + "kind": "convolution" + }, + { + "input": "states_21", + "output": "slice_100", + "kind": "recurrent" + }, + { + "input": "states_22", + "output": "cache_update_4", + "kind": "kv" + }, + { + "input": "states_23", + "output": "cache_update_5", + "kind": "kv" + }, + { + "input": "states_24", + "output": "slice_112", + "kind": "convolution" + }, + { + "input": "states_25", + "output": "slice_118", + "kind": "recurrent" + }, + { + "input": "states_26", + "output": "slice_120", + "kind": "convolution" + }, + { + "input": "states_27", + "output": "slice_126", + "kind": "recurrent" + }, + { + "input": "states_28", + "output": "slice_128", + "kind": "convolution" + }, + { + "input": "states_29", + "output": "slice_134", + "kind": "recurrent" + }, + { + "input": "states_30", + "output": "cache_update_6", + "kind": "kv" + }, + { + "input": "states_31", + "output": "cache_update_7", + "kind": "kv" + }, + { + "input": "states_32", + "output": "slice_146", + "kind": "convolution" + }, + { + "input": "states_33", + "output": "slice_152", + "kind": "recurrent" + }, + { + "input": "states_34", + "output": "slice_154", + "kind": "convolution" + }, + { + "input": "states_35", + "output": "slice_160", + "kind": "recurrent" + }, + { + "input": "states_36", + "output": "slice_162", + "kind": "convolution" + }, + { + "input": "states_37", + "output": "slice_168", + "kind": "recurrent" + }, + { + "input": "states_38", + "output": "cache_update_8", + "kind": "kv" + }, + { + "input": "states_39", + "output": "cache_update_9", + "kind": "kv" + }, + { + "input": "states_40", + "output": "slice_180", + "kind": "convolution" + }, + { + "input": "states_41", + "output": "slice_186", + "kind": "recurrent" + }, + { + "input": "states_42", + "output": "slice_188", + "kind": "convolution" + }, + { + "input": "states_43", + "output": "slice_194", + "kind": "recurrent" + }, + { + "input": "states_44", + "output": "slice_196", + "kind": "convolution" + }, + { + "input": "states_45", + "output": "slice_202", + "kind": "recurrent" + }, + { + "input": "states_46", + "output": "cache_update_10", + "kind": "kv" + }, + { + "input": "states_47", + "output": "cache_update_11", + "kind": "kv" + } + ], + "dispatches": [ + { + "node": "embedding__part_0", + "inputs": [ + "input_ids", + "p_model_model_language_model_embed_tokens_weight__rows_0_131072", + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072" + ], + "output": "embedding__part_0", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/7dfc17a30d9796026977dc65bedef251d9b2b5fc7eea0ebee019ad956226a513.wgsl" + }, + { + "node": "embedding__part_1", + "inputs": [ + "input_ids", + "p_model_model_language_model_embed_tokens_weight__rows_131072_117248", + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248" + ], + "output": "embedding__part_1", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/2ed0344f526ea256cc4b9261d44a1223b139d0294a07ee072107dc85fa6a4e0f.wgsl" + }, + { + "node": "embedding", + "inputs": [ + "embedding__part_0", + "embedding__part_1" + ], + "output": "embedding", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "unsqueeze", + "inputs": [ + "position_ids" + ], + "output": "unsqueeze", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/e867b978e7008207276fa1603c3c1c968a91d5b02e70daa63b1c9e462b6c36d2.wgsl" + }, + { + "node": "expand", + "inputs": [ + "unsqueeze" + ], + "output": "expand", + "workgroups": [ + 4, + 1, + 1 + ], + "kernel": "kernels/a896c93b04922c0f78f1c77fbaabbc79eb71776223748bb32a6d5339e32029e8.wgsl" + }, + { + "node": "slice_1", + "inputs": [ + "expand" + ], + "output": "slice_1", + "workgroups": [ + 3, + 1, + 1 + ], + "kernel": "kernels/71d38690bcc7fc30f01f1c8fa2f13a38beff889fcb3ff387861bb70038e92129.wgsl" + }, + { + "node": "expand_1", + "inputs": [ + "b_model_model_language_model_rotary_emb_inv_freq" + ], + "output": "expand_1", + "workgroups": [ + 2, + 1, + 1 + ], + "kernel": "kernels/a77f14c89662d2d3da8abd8f109667793dc662d307e5e1c61123be753e25437b.wgsl" + }, + { + "node": "_to_copy", + "inputs": [ + "slice_1" + ], + "output": "_to_copy", + "workgroups": [ + 3, + 1, + 1 + ], + "kernel": "kernels/da7c63a84d7f2c46e8234a34c0905cfad6229740145070df00819509f8ab39de.wgsl" + }, + { + "node": "matmul", + "inputs": [ + "expand_1", + "_to_copy" + ], + "output": "matmul", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/dd544e5a005b68617e9a3557e1d24dd6282a5a62f47cae4cc41b619d3a1cd603.wgsl" + }, + { + "node": "transpose", + "inputs": [ + "matmul" + ], + "output": "transpose", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/84101fc314579b90d71da4cca3a2c2cdc2074d96d2861c4b8e38c22cc1903138.wgsl" + }, + { + "node": "select", + "inputs": [ + "transpose" + ], + "output": "select", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/d8757289e027d60366190dda56db53f6d6a3358fd813e84ca58f2076ddb0bf3e.wgsl" + }, + { + "node": "select_1", + "inputs": [ + "transpose" + ], + "output": "select_1", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/f5f6f056866d42771d34664166a72300424224e9d6bd1d57de1805e7476a2050.wgsl" + }, + { + "node": "slice_2", + "inputs": [ + "select_1" + ], + "output": "slice_2", + "workgroups": [ + 11, + 1, + 1 + ], + "kernel": "kernels/efcecdf33b6927a4add3f254d95cad6d18d755d65e3f4873e8b7f5c057646a9b.wgsl" + }, + { + "node": "slice_3", + "inputs": [ + "select" + ], + "output": "slice_3", + "workgroups": [ + 11, + 1, + 1 + ], + "kernel": "kernels/efcecdf33b6927a4add3f254d95cad6d18d755d65e3f4873e8b7f5c057646a9b.wgsl" + }, + { + "node": "copy", + "inputs": [ + "slice_2" + ], + "output": "copy", + "workgroups": [ + 11, + 1, + 1 + ], + "kernel": "kernels/48608b2b7bfbe0b65d9e2d82a94a76683315421acb1bf168b4baf14a3c139f76.wgsl" + }, + { + "node": "transpose_1", + "inputs": [ + "matmul" + ], + "output": "transpose_1", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/84101fc314579b90d71da4cca3a2c2cdc2074d96d2861c4b8e38c22cc1903138.wgsl" + }, + { + "node": "select_2", + "inputs": [ + "transpose_1" + ], + "output": "select_2", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/d8757289e027d60366190dda56db53f6d6a3358fd813e84ca58f2076ddb0bf3e.wgsl" + }, + { + "node": "slice_scatter", + "inputs": [ + "copy", + "select_2" + ], + "output": "slice_scatter", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/9f040facf0594804dba5b0f10f53b11b7223a1e1705896f95a8f95aa0fff5eec.wgsl" + }, + { + "node": "select_scatter", + "inputs": [ + "slice_scatter", + "transpose_1" + ], + "output": "select_scatter", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/691af757595593839af5e5e97ee656472f6e7c7406f57d1d9faae74816d9fc91.wgsl" + }, + { + "node": "transpose_2", + "inputs": [ + "select_scatter" + ], + "output": "transpose_2", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/06eecde176beab9ff488ad533f2f4513e09bc49a30c4358f9648076f619f9992.wgsl" + }, + { + "node": "transpose_7", + "inputs": [ + "transpose_2" + ], + "output": "transpose_7", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/84101fc314579b90d71da4cca3a2c2cdc2074d96d2861c4b8e38c22cc1903138.wgsl" + }, + { + "node": "select_7", + "inputs": [ + "transpose_7" + ], + "output": "select_7", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/d8757289e027d60366190dda56db53f6d6a3358fd813e84ca58f2076ddb0bf3e.wgsl" + }, + { + "node": "slice_7", + "inputs": [ + "select_7" + ], + "output": "slice_7", + "workgroups": [ + 10, + 1, + 1 + ], + "kernel": "kernels/fb200142bcbd435ed588a9fc45924edb21845ff3ab0ddb0271c5d5d03adf04bb.wgsl" + }, + { + "node": "transpose_8", + "inputs": [ + "transpose_2" + ], + "output": "transpose_8", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/84101fc314579b90d71da4cca3a2c2cdc2074d96d2861c4b8e38c22cc1903138.wgsl" + }, + { + "node": "select_8", + "inputs": [ + "transpose_8" + ], + "output": "select_8", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/c83a2a6faf563ec3bbad43a3e108d58c09cd8aeed01b38e4e82d1c3c9b650c21.wgsl" + }, + { + "node": "slice_8", + "inputs": [ + "select_8" + ], + "output": "slice_8", + "workgroups": [ + 10, + 1, + 1 + ], + "kernel": "kernels/fb200142bcbd435ed588a9fc45924edb21845ff3ab0ddb0271c5d5d03adf04bb.wgsl" + }, + { + "node": "copy_1", + "inputs": [ + "slice_8" + ], + "output": "copy_1", + "workgroups": [ + 10, + 1, + 1 + ], + "kernel": "kernels/56167a7e8fb1c67fa72948a8baa71866d54cc775be83f8cfa75a947ac0f58b12.wgsl" + }, + { + "node": "transpose_9", + "inputs": [ + "transpose_2" + ], + "output": "transpose_9", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/84101fc314579b90d71da4cca3a2c2cdc2074d96d2861c4b8e38c22cc1903138.wgsl" + }, + { + "node": "select_9", + "inputs": [ + "transpose_9" + ], + "output": "select_9", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/d8757289e027d60366190dda56db53f6d6a3358fd813e84ca58f2076ddb0bf3e.wgsl" + }, + { + "node": "slice_scatter_1", + "inputs": [ + "copy_1", + "select_9" + ], + "output": "slice_scatter_1", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/b3bb2905555188f535b6b8bc1d050c142b5fea8222d9ef193466f80aa2c54dea.wgsl" + }, + { + "node": "select_scatter_1", + "inputs": [ + "slice_scatter_1", + "transpose_9" + ], + "output": "select_scatter_1", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/691af757595593839af5e5e97ee656472f6e7c7406f57d1d9faae74816d9fc91.wgsl" + }, + { + "node": "transpose_10", + "inputs": [ + "select_scatter_1" + ], + "output": "transpose_10", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/06eecde176beab9ff488ad533f2f4513e09bc49a30c4358f9648076f619f9992.wgsl" + }, + { + "node": "transpose_12", + "inputs": [ + "transpose_10" + ], + "output": "transpose_12", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/84101fc314579b90d71da4cca3a2c2cdc2074d96d2861c4b8e38c22cc1903138.wgsl" + }, + { + "node": "select_11", + "inputs": [ + "transpose_12" + ], + "output": "select_11", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/d8757289e027d60366190dda56db53f6d6a3358fd813e84ca58f2076ddb0bf3e.wgsl" + }, + { + "node": "cat", + "inputs": [ + "select_11" + ], + "output": "cat", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/0337d8a473b6e485bd383533df7d373dbbfe7a5f3a982424004560b7213fbaa3.wgsl" + }, + { + "node": "cos", + "inputs": [ + "cat" + ], + "output": "cos", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/7adea2b745b20a2abb96d02a1fca6db5f06f3c5aac850cf8ae1afa6f50868ded.wgsl" + }, + { + "node": "mul", + "inputs": [ + "cos" + ], + "output": "mul", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/2b9e4b05ef5623d721dbe56b5812e551efcc6fd833d81064320e86fe94d8c670.wgsl" + }, + { + "node": "sin", + "inputs": [ + "cat" + ], + "output": "sin", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/28119aa01122c42bd26ceb37129fe0c95b2a9192163f6dabd90fe058b1e14c54.wgsl" + }, + { + "node": "mul_1", + "inputs": [ + "sin" + ], + "output": "mul_1", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/2b9e4b05ef5623d721dbe56b5812e551efcc6fd833d81064320e86fe94d8c670.wgsl" + }, + { + "node": "mul_3", + "inputs": [ + "embedding", + "p_model_model_language_model_layers_0_input_layernorm_weight" + ], + "output": "mul_3", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear", + "inputs": [ + "mul_3", + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_13", + "inputs": [ + "linear" + ], + "output": "transpose_13", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_1", + "inputs": [ + "mul_3", + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_1", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_2", + "inputs": [ + "mul_3", + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_3", + "inputs": [ + "mul_3", + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_1", + "inputs": [ + "states_0", + "transpose_13" + ], + "output": "cat_1", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_10", + "inputs": [ + "cat_1" + ], + "output": "slice_10", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_0" + }, + { + "node": "conv1d", + "inputs": [ + "cat_1", + "p_model_model_language_model_layers_0_linear_attn_conv1d_weight" + ], + "output": "conv1d", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_11", + "inputs": [ + "conv1d" + ], + "output": "slice_11", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu", + "inputs": [ + "slice_11" + ], + "output": "silu", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_12", + "inputs": [ + "silu" + ], + "output": "slice_12", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_14", + "inputs": [ + "slice_12" + ], + "output": "transpose_14", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_13", + "inputs": [ + "transpose_14" + ], + "output": "slice_13", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_14", + "inputs": [ + "transpose_14" + ], + "output": "slice_14", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_15", + "inputs": [ + "transpose_14" + ], + "output": "slice_15", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid", + "inputs": [ + "linear_2" + ], + "output": "sigmoid", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp", + "inputs": [ + "p_model_model_language_model_layers_0_linear_attn_a_log" + ], + "output": "exp", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg", + "inputs": [ + "exp" + ], + "output": "neg", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_2", + "inputs": [ + "linear_3", + "p_model_model_language_model_layers_0_linear_attn_dt_bias" + ], + "output": "add_2", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus", + "inputs": [ + "add_2" + ], + "output": "softplus", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_4", + "inputs": [ + "neg", + "softplus" + ], + "output": "mul_4", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_5", + "inputs": [ + "slice_13" + ], + "output": "mul_5", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_1", + "inputs": [ + "mul_5" + ], + "output": "sum_1", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_3", + "inputs": [ + "sum_1" + ], + "output": "add_3", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_1", + "inputs": [ + "add_3" + ], + "output": "rsqrt_1", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_6", + "inputs": [ + "slice_13", + "rsqrt_1" + ], + "output": "mul_6", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_7", + "inputs": [ + "slice_14" + ], + "output": "mul_7", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_2", + "inputs": [ + "mul_7" + ], + "output": "sum_2", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_4", + "inputs": [ + "sum_2" + ], + "output": "add_4", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_2", + "inputs": [ + "add_4" + ], + "output": "rsqrt_2", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_8", + "inputs": [ + "slice_14", + "rsqrt_2" + ], + "output": "mul_8", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule", + "inputs": [ + "mul_6", + "mul_8", + "slice_15", + "mul_4", + "sigmoid", + "states_1" + ], + "output": "normalized_gated_delta_rule", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_16", + "inputs": [ + "normalized_gated_delta_rule" + ], + "output": "slice_16", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_1" + }, + { + "node": "slice_17", + "inputs": [ + "normalized_gated_delta_rule" + ], + "output": "slice_17", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_15", + "inputs": [ + "slice_17" + ], + "output": "transpose_15", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_10", + "inputs": [ + "transpose_15", + "p_model_model_language_model_layers_0_linear_attn_norm_weight" + ], + "output": "mul_10", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_11", + "inputs": [ + "linear_1", + "mul_10" + ], + "output": "mul_11", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_4", + "inputs": [ + "mul_11", + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_4", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_6", + "inputs": [ + "embedding", + "linear_4" + ], + "output": "add_6", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_13", + "inputs": [ + "add_6", + "p_model_model_language_model_layers_0_post_attention_layernorm_weight" + ], + "output": "mul_13", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_5", + "inputs": [ + "mul_13", + "p_model_model_language_model_layers_0_mlp_gate_proj_weight", + "p_model_model_language_model_layers_0_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_5", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_6", + "inputs": [ + "mul_13", + "p_model_model_language_model_layers_0_mlp_up_proj_weight", + "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_6", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_14", + "inputs": [ + "linear_5", + "linear_6" + ], + "output": "mul_14", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_7", + "inputs": [ + "mul_14", + "p_model_model_language_model_layers_0_mlp_down_proj_weight", + "p_model_model_language_model_layers_0_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_7", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_9", + "inputs": [ + "add_6", + "linear_7" + ], + "output": "add_9", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_16", + "inputs": [ + "add_9", + "p_model_model_language_model_layers_1_input_layernorm_weight" + ], + "output": "mul_16", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_8", + "inputs": [ + "mul_16", + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_8", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_16", + "inputs": [ + "linear_8" + ], + "output": "transpose_16", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_9", + "inputs": [ + "mul_16", + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_9", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_10", + "inputs": [ + "mul_16", + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_11", + "inputs": [ + "mul_16", + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_2", + "inputs": [ + "states_2", + "transpose_16" + ], + "output": "cat_2", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_18", + "inputs": [ + "cat_2" + ], + "output": "slice_18", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_2" + }, + { + "node": "conv1d_1", + "inputs": [ + "cat_2", + "p_model_model_language_model_layers_1_linear_attn_conv1d_weight" + ], + "output": "conv1d_1", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_19", + "inputs": [ + "conv1d_1" + ], + "output": "slice_19", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_3", + "inputs": [ + "slice_19" + ], + "output": "silu_3", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_20", + "inputs": [ + "silu_3" + ], + "output": "slice_20", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_17", + "inputs": [ + "slice_20" + ], + "output": "transpose_17", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_21", + "inputs": [ + "transpose_17" + ], + "output": "slice_21", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_22", + "inputs": [ + "transpose_17" + ], + "output": "slice_22", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_23", + "inputs": [ + "transpose_17" + ], + "output": "slice_23", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_1", + "inputs": [ + "linear_10" + ], + "output": "sigmoid_1", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_1", + "inputs": [ + "p_model_model_language_model_layers_1_linear_attn_a_log" + ], + "output": "exp_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_1", + "inputs": [ + "exp_1" + ], + "output": "neg_1", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_12", + "inputs": [ + "linear_11", + "p_model_model_language_model_layers_1_linear_attn_dt_bias" + ], + "output": "add_12", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_1", + "inputs": [ + "add_12" + ], + "output": "softplus_1", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_17", + "inputs": [ + "neg_1", + "softplus_1" + ], + "output": "mul_17", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_18", + "inputs": [ + "slice_21" + ], + "output": "mul_18", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_3", + "inputs": [ + "mul_18" + ], + "output": "sum_3", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_13", + "inputs": [ + "sum_3" + ], + "output": "add_13", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_6", + "inputs": [ + "add_13" + ], + "output": "rsqrt_6", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_19", + "inputs": [ + "slice_21", + "rsqrt_6" + ], + "output": "mul_19", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_20", + "inputs": [ + "slice_22" + ], + "output": "mul_20", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_4", + "inputs": [ + "mul_20" + ], + "output": "sum_4", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_14", + "inputs": [ + "sum_4" + ], + "output": "add_14", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_7", + "inputs": [ + "add_14" + ], + "output": "rsqrt_7", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_21", + "inputs": [ + "slice_22", + "rsqrt_7" + ], + "output": "mul_21", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_1", + "inputs": [ + "mul_19", + "mul_21", + "slice_23", + "mul_17", + "sigmoid_1", + "states_3" + ], + "output": "normalized_gated_delta_rule_1", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_24", + "inputs": [ + "normalized_gated_delta_rule_1" + ], + "output": "slice_24", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_3" + }, + { + "node": "slice_25", + "inputs": [ + "normalized_gated_delta_rule_1" + ], + "output": "slice_25", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_18", + "inputs": [ + "slice_25" + ], + "output": "transpose_18", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_23", + "inputs": [ + "transpose_18", + "p_model_model_language_model_layers_1_linear_attn_norm_weight" + ], + "output": "mul_23", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_24", + "inputs": [ + "linear_9", + "mul_23" + ], + "output": "mul_24", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_12", + "inputs": [ + "mul_24", + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_12", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_16", + "inputs": [ + "add_9", + "linear_12" + ], + "output": "add_16", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_26", + "inputs": [ + "add_16", + "p_model_model_language_model_layers_1_post_attention_layernorm_weight" + ], + "output": "mul_26", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_13", + "inputs": [ + "mul_26", + "p_model_model_language_model_layers_1_mlp_gate_proj_weight", + "p_model_model_language_model_layers_1_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_13", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_14", + "inputs": [ + "mul_26", + "p_model_model_language_model_layers_1_mlp_up_proj_weight", + "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_14", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_27", + "inputs": [ + "linear_13", + "linear_14" + ], + "output": "mul_27", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_15", + "inputs": [ + "mul_27", + "p_model_model_language_model_layers_1_mlp_down_proj_weight", + "p_model_model_language_model_layers_1_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_15", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_19", + "inputs": [ + "add_16", + "linear_15" + ], + "output": "add_19", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_29", + "inputs": [ + "add_19", + "p_model_model_language_model_layers_2_input_layernorm_weight" + ], + "output": "mul_29", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_16", + "inputs": [ + "mul_29", + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_16", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_19", + "inputs": [ + "linear_16" + ], + "output": "transpose_19", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_17", + "inputs": [ + "mul_29", + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_17", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_18", + "inputs": [ + "mul_29", + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_18", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_19", + "inputs": [ + "mul_29", + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_19", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_3", + "inputs": [ + "states_4", + "transpose_19" + ], + "output": "cat_3", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_26", + "inputs": [ + "cat_3" + ], + "output": "slice_26", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_4" + }, + { + "node": "conv1d_2", + "inputs": [ + "cat_3", + "p_model_model_language_model_layers_2_linear_attn_conv1d_weight" + ], + "output": "conv1d_2", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_27", + "inputs": [ + "conv1d_2" + ], + "output": "slice_27", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_6", + "inputs": [ + "slice_27" + ], + "output": "silu_6", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_28", + "inputs": [ + "silu_6" + ], + "output": "slice_28", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_20", + "inputs": [ + "slice_28" + ], + "output": "transpose_20", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_29", + "inputs": [ + "transpose_20" + ], + "output": "slice_29", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_30", + "inputs": [ + "transpose_20" + ], + "output": "slice_30", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_31", + "inputs": [ + "transpose_20" + ], + "output": "slice_31", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_2", + "inputs": [ + "linear_18" + ], + "output": "sigmoid_2", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_2", + "inputs": [ + "p_model_model_language_model_layers_2_linear_attn_a_log" + ], + "output": "exp_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_2", + "inputs": [ + "exp_2" + ], + "output": "neg_2", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_22", + "inputs": [ + "linear_19", + "p_model_model_language_model_layers_2_linear_attn_dt_bias" + ], + "output": "add_22", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_2", + "inputs": [ + "add_22" + ], + "output": "softplus_2", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_30", + "inputs": [ + "neg_2", + "softplus_2" + ], + "output": "mul_30", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_31", + "inputs": [ + "slice_29" + ], + "output": "mul_31", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_5", + "inputs": [ + "mul_31" + ], + "output": "sum_5", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_23", + "inputs": [ + "sum_5" + ], + "output": "add_23", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_11", + "inputs": [ + "add_23" + ], + "output": "rsqrt_11", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_32", + "inputs": [ + "slice_29", + "rsqrt_11" + ], + "output": "mul_32", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_33", + "inputs": [ + "slice_30" + ], + "output": "mul_33", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_6", + "inputs": [ + "mul_33" + ], + "output": "sum_6", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_24", + "inputs": [ + "sum_6" + ], + "output": "add_24", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_12", + "inputs": [ + "add_24" + ], + "output": "rsqrt_12", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_34", + "inputs": [ + "slice_30", + "rsqrt_12" + ], + "output": "mul_34", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_2", + "inputs": [ + "mul_32", + "mul_34", + "slice_31", + "mul_30", + "sigmoid_2", + "states_5" + ], + "output": "normalized_gated_delta_rule_2", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_32", + "inputs": [ + "normalized_gated_delta_rule_2" + ], + "output": "slice_32", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_5" + }, + { + "node": "slice_33", + "inputs": [ + "normalized_gated_delta_rule_2" + ], + "output": "slice_33", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_21", + "inputs": [ + "slice_33" + ], + "output": "transpose_21", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_36", + "inputs": [ + "transpose_21", + "p_model_model_language_model_layers_2_linear_attn_norm_weight" + ], + "output": "mul_36", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_37", + "inputs": [ + "linear_17", + "mul_36" + ], + "output": "mul_37", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_20", + "inputs": [ + "mul_37", + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_20", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_26", + "inputs": [ + "add_19", + "linear_20" + ], + "output": "add_26", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_39", + "inputs": [ + "add_26", + "p_model_model_language_model_layers_2_post_attention_layernorm_weight" + ], + "output": "mul_39", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_21", + "inputs": [ + "mul_39", + "p_model_model_language_model_layers_2_mlp_gate_proj_weight", + "p_model_model_language_model_layers_2_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_21", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_22", + "inputs": [ + "mul_39", + "p_model_model_language_model_layers_2_mlp_up_proj_weight", + "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_22", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_40", + "inputs": [ + "linear_21", + "linear_22" + ], + "output": "mul_40", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_23", + "inputs": [ + "mul_40", + "p_model_model_language_model_layers_2_mlp_down_proj_weight", + "p_model_model_language_model_layers_2_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_23", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_29", + "inputs": [ + "add_26", + "linear_23" + ], + "output": "add_29", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_42", + "inputs": [ + "add_29", + "p_model_model_language_model_layers_3_input_layernorm_weight" + ], + "output": "mul_42", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_24", + "inputs": [ + "mul_42", + "p_model_model_language_model_layers_3_self_attn_q_proj_weight", + "p_model_model_language_model_layers_3_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_24", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d1e9766c981b6c437331275b2afdfc74e0789005b52c2015839fc2394961da1b.wgsl" + }, + { + "node": "slice_34", + "inputs": [ + "linear_24" + ], + "output": "slice_34", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3a76085983a5330bf07a4166ea4975700403f9d8035a8ef450f7f18b9cf47c66.wgsl" + }, + { + "node": "slice_35", + "inputs": [ + "linear_24" + ], + "output": "slice_35", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9044193f1f4724f15ef5fb41040b7fbac5f62fcaea80594737214bf9cd6be41a.wgsl" + }, + { + "node": "mul_44", + "inputs": [ + "slice_34", + "p_model_model_language_model_layers_3_self_attn_q_norm_weight" + ], + "output": "mul_44", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/ef364f47cced448917d1b6842c708dee101d53d7a781014f9a822214cea72bf8.wgsl" + }, + { + "node": "transpose_22", + "inputs": [ + "mul_44" + ], + "output": "transpose_22", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/1d0cbb068fe17123e646934bfa1259f19e5db04e45d72a157d3ea4fa48c81aa0.wgsl" + }, + { + "node": "linear_25", + "inputs": [ + "mul_42", + "p_model_model_language_model_layers_3_self_attn_k_proj_weight", + "p_model_model_language_model_layers_3_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_25", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/c5b7564d87c89606da1cba30e05e401e4fb8c83957e779a694d2ec2c472d88ad.wgsl" + }, + { + "node": "mul_46", + "inputs": [ + "linear_25", + "p_model_model_language_model_layers_3_self_attn_k_norm_weight" + ], + "output": "mul_46", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl" + }, + { + "node": "transpose_23", + "inputs": [ + "mul_46" + ], + "output": "transpose_23", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/bcb78b92dbbfeb212c190e064bc2f9d8e79c201398cf38c9aa62df6fdba94469.wgsl" + }, + { + "node": "linear_26", + "inputs": [ + "mul_42", + "p_model_model_language_model_layers_3_self_attn_v_proj_weight", + "p_model_model_language_model_layers_3_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_26", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/c5b7564d87c89606da1cba30e05e401e4fb8c83957e779a694d2ec2c472d88ad.wgsl" + }, + { + "node": "transpose_24", + "inputs": [ + "linear_26" + ], + "output": "transpose_24", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/bcb78b92dbbfeb212c190e064bc2f9d8e79c201398cf38c9aa62df6fdba94469.wgsl" + }, + { + "node": "slice_36", + "inputs": [ + "transpose_22" + ], + "output": "slice_36", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/4d8ebffbbf193b18c13a5cc17d286bf5c92c5ba53ca022af78a9151ec1e5b9c0.wgsl" + }, + { + "node": "slice_37", + "inputs": [ + "transpose_22" + ], + "output": "slice_37", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/4406ab1e6e4b9eeb91ace297320aa05a55cc481e02894a4a74b655b4070ade36.wgsl" + }, + { + "node": "slice_38", + "inputs": [ + "transpose_23" + ], + "output": "slice_38", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/89f98b79c18e7f4bbbd79cbd8d3e80108d0e1e4407c161d9637b6a88d3fbe06b.wgsl" + }, + { + "node": "slice_39", + "inputs": [ + "transpose_23" + ], + "output": "slice_39", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd70e4fc42ac55f0f52612841bcc3f9fe7df5bbb0fc76f94e56a131a92454658.wgsl" + }, + { + "node": "mul_47", + "inputs": [ + "slice_36", + "mul" + ], + "output": "mul_47", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl" + }, + { + "node": "slice_40", + "inputs": [ + "slice_36" + ], + "output": "slice_40", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/2c66dcac8fdd3caa9ff1e2441524070c21c24df9526bee44c49e765a8d4c8222.wgsl" + }, + { + "node": "slice_41", + "inputs": [ + "slice_36" + ], + "output": "slice_41", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/33e8ec04c66adb4988f562f0f9d3e92d29dfda91a56d0ed62be598c7c737f5ba.wgsl" + }, + { + "node": "neg_3", + "inputs": [ + "slice_41" + ], + "output": "neg_3", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/3e34409407100d4858c4c109cf62f3eaf2f4c6081166b0ec8679105e6dfbe7fe.wgsl" + }, + { + "node": "cat_4", + "inputs": [ + "neg_3", + "slice_40" + ], + "output": "cat_4", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/85964f8bbe5a2e4d430930e1050ea76871f0f2453c3cd4ba020002c0562ca680.wgsl" + }, + { + "node": "mul_48", + "inputs": [ + "cat_4", + "mul_1" + ], + "output": "mul_48", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl" + }, + { + "node": "add_36", + "inputs": [ + "mul_47", + "mul_48" + ], + "output": "add_36", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/d2866fd21f58839ea7d43bf83fde4243c38e4f1771034520abb020be5d819a96.wgsl" + }, + { + "node": "mul_49", + "inputs": [ + "slice_38", + "mul" + ], + "output": "mul_49", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl" + }, + { + "node": "slice_42", + "inputs": [ + "slice_38" + ], + "output": "slice_42", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/7b7c42c04ea3c6695ddead641e88b82d8f130015e138fc323eec509b5b56cfa9.wgsl" + }, + { + "node": "slice_43", + "inputs": [ + "slice_38" + ], + "output": "slice_43", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/6d82b7ed99a04f03fa0b9d241188d9613fa3ce9ff425cf8ee30a7f4ce6ded13d.wgsl" + }, + { + "node": "neg_4", + "inputs": [ + "slice_43" + ], + "output": "neg_4", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/adf1e58a124e1301625024ae142dca5fb1dd50799fa94b52eaeffb82ea99852c.wgsl" + }, + { + "node": "cat_5", + "inputs": [ + "neg_4", + "slice_42" + ], + "output": "cat_5", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl" + }, + { + "node": "mul_50", + "inputs": [ + "cat_5", + "mul_1" + ], + "output": "mul_50", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl" + }, + { + "node": "add_37", + "inputs": [ + "mul_49", + "mul_50" + ], + "output": "add_37", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl" + }, + { + "node": "cat_6", + "inputs": [ + "add_36", + "slice_37" + ], + "output": "cat_6", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/8affc7c0e85127430e406be35ebb5c5354a8dd26ecc4ae9ab6ace80ce6775f46.wgsl" + }, + { + "node": "cat_7", + "inputs": [ + "add_37", + "slice_39" + ], + "output": "cat_7", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl" + }, + { + "node": "cache_update", + "inputs": [ + "cat_7", + "position_ids" + ], + "output": "cache_update", + "workgroups": [ + 512, + 1, + 1 + ], + "inPlaceOf": "states_6", + "kernel": "kernels/e369b5c120d84bfebbc35e3da096c33e55d20a607aecd5d82da5de4c95176e00.wgsl" + }, + { + "node": "cache_update_1", + "inputs": [ + "transpose_24", + "position_ids" + ], + "output": "cache_update_1", + "workgroups": [ + 512, + 1, + 1 + ], + "inPlaceOf": "states_7", + "kernel": "kernels/e369b5c120d84bfebbc35e3da096c33e55d20a607aecd5d82da5de4c95176e00.wgsl" + }, + { + "node": "matmul_2", + "inputs": [ + "cat_6", + "cache_update", + "cache_update_1", + "attention_mask" + ], + "output": "matmul_2", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/140ebe12d31f1420a8a159b91ae1773390190f496e06e0e5d7b5351f38245068.wgsl" + }, + { + "node": "transpose_26", + "inputs": [ + "matmul_2" + ], + "output": "transpose_26", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4f2c0d08150d4b97d90c244db2d41baca81484c2630dbbd6d586f2d8a4b07b8b.wgsl" + }, + { + "node": "sigmoid_3", + "inputs": [ + "slice_35" + ], + "output": "sigmoid_3", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/50535e4c99d9fb4a272d8260a1415f6b38a7ad1025da64e145f230df697d9b73.wgsl" + }, + { + "node": "mul_52", + "inputs": [ + "transpose_26", + "sigmoid_3" + ], + "output": "mul_52", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/a9f6bc361f1ea31f79f0ef6164872eeb04896a45338871a2d779789b47b1920b.wgsl" + }, + { + "node": "linear_27", + "inputs": [ + "mul_52", + "p_model_model_language_model_layers_3_self_attn_o_proj_weight", + "p_model_model_language_model_layers_3_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_27", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_39", + "inputs": [ + "add_29", + "linear_27" + ], + "output": "add_39", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_54", + "inputs": [ + "add_39", + "p_model_model_language_model_layers_3_post_attention_layernorm_weight" + ], + "output": "mul_54", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_28", + "inputs": [ + "mul_54", + "p_model_model_language_model_layers_3_mlp_gate_proj_weight", + "p_model_model_language_model_layers_3_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_28", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_29", + "inputs": [ + "mul_54", + "p_model_model_language_model_layers_3_mlp_up_proj_weight", + "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_29", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_55", + "inputs": [ + "linear_28", + "linear_29" + ], + "output": "mul_55", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_30", + "inputs": [ + "mul_55", + "p_model_model_language_model_layers_3_mlp_down_proj_weight", + "p_model_model_language_model_layers_3_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_30", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_42", + "inputs": [ + "add_39", + "linear_30" + ], + "output": "add_42", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_57", + "inputs": [ + "add_42", + "p_model_model_language_model_layers_4_input_layernorm_weight" + ], + "output": "mul_57", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_31", + "inputs": [ + "mul_57", + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_31", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_27", + "inputs": [ + "linear_31" + ], + "output": "transpose_27", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_32", + "inputs": [ + "mul_57", + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_32", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_33", + "inputs": [ + "mul_57", + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_33", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_34", + "inputs": [ + "mul_57", + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_34", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_8", + "inputs": [ + "states_8", + "transpose_27" + ], + "output": "cat_8", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_44", + "inputs": [ + "cat_8" + ], + "output": "slice_44", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_8" + }, + { + "node": "conv1d_3", + "inputs": [ + "cat_8", + "p_model_model_language_model_layers_4_linear_attn_conv1d_weight" + ], + "output": "conv1d_3", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_45", + "inputs": [ + "conv1d_3" + ], + "output": "slice_45", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_10", + "inputs": [ + "slice_45" + ], + "output": "silu_10", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_46", + "inputs": [ + "silu_10" + ], + "output": "slice_46", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_28", + "inputs": [ + "slice_46" + ], + "output": "transpose_28", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_47", + "inputs": [ + "transpose_28" + ], + "output": "slice_47", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_48", + "inputs": [ + "transpose_28" + ], + "output": "slice_48", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_49", + "inputs": [ + "transpose_28" + ], + "output": "slice_49", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_4", + "inputs": [ + "linear_33" + ], + "output": "sigmoid_4", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_3", + "inputs": [ + "p_model_model_language_model_layers_4_linear_attn_a_log" + ], + "output": "exp_3", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_5", + "inputs": [ + "exp_3" + ], + "output": "neg_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_45", + "inputs": [ + "linear_34", + "p_model_model_language_model_layers_4_linear_attn_dt_bias" + ], + "output": "add_45", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_3", + "inputs": [ + "add_45" + ], + "output": "softplus_3", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_58", + "inputs": [ + "neg_5", + "softplus_3" + ], + "output": "mul_58", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_59", + "inputs": [ + "slice_47" + ], + "output": "mul_59", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_7", + "inputs": [ + "mul_59" + ], + "output": "sum_7", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_46", + "inputs": [ + "sum_7" + ], + "output": "add_46", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_20", + "inputs": [ + "add_46" + ], + "output": "rsqrt_20", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_60", + "inputs": [ + "slice_47", + "rsqrt_20" + ], + "output": "mul_60", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_61", + "inputs": [ + "slice_48" + ], + "output": "mul_61", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_8", + "inputs": [ + "mul_61" + ], + "output": "sum_8", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_47", + "inputs": [ + "sum_8" + ], + "output": "add_47", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_21", + "inputs": [ + "add_47" + ], + "output": "rsqrt_21", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_62", + "inputs": [ + "slice_48", + "rsqrt_21" + ], + "output": "mul_62", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_3", + "inputs": [ + "mul_60", + "mul_62", + "slice_49", + "mul_58", + "sigmoid_4", + "states_9" + ], + "output": "normalized_gated_delta_rule_3", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_50", + "inputs": [ + "normalized_gated_delta_rule_3" + ], + "output": "slice_50", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_9" + }, + { + "node": "slice_51", + "inputs": [ + "normalized_gated_delta_rule_3" + ], + "output": "slice_51", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_29", + "inputs": [ + "slice_51" + ], + "output": "transpose_29", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_64", + "inputs": [ + "transpose_29", + "p_model_model_language_model_layers_4_linear_attn_norm_weight" + ], + "output": "mul_64", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_65", + "inputs": [ + "linear_32", + "mul_64" + ], + "output": "mul_65", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_35", + "inputs": [ + "mul_65", + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_35", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_49", + "inputs": [ + "add_42", + "linear_35" + ], + "output": "add_49", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_67", + "inputs": [ + "add_49", + "p_model_model_language_model_layers_4_post_attention_layernorm_weight" + ], + "output": "mul_67", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_36", + "inputs": [ + "mul_67", + "p_model_model_language_model_layers_4_mlp_gate_proj_weight", + "p_model_model_language_model_layers_4_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_36", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_37", + "inputs": [ + "mul_67", + "p_model_model_language_model_layers_4_mlp_up_proj_weight", + "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_37", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_68", + "inputs": [ + "linear_36", + "linear_37" + ], + "output": "mul_68", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_38", + "inputs": [ + "mul_68", + "p_model_model_language_model_layers_4_mlp_down_proj_weight", + "p_model_model_language_model_layers_4_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_38", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_52", + "inputs": [ + "add_49", + "linear_38" + ], + "output": "add_52", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_70", + "inputs": [ + "add_52", + "p_model_model_language_model_layers_5_input_layernorm_weight" + ], + "output": "mul_70", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_39", + "inputs": [ + "mul_70", + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_39", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_30", + "inputs": [ + "linear_39" + ], + "output": "transpose_30", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_40", + "inputs": [ + "mul_70", + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_40", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_41", + "inputs": [ + "mul_70", + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_41", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_42", + "inputs": [ + "mul_70", + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_42", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_9", + "inputs": [ + "states_10", + "transpose_30" + ], + "output": "cat_9", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_52", + "inputs": [ + "cat_9" + ], + "output": "slice_52", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_10" + }, + { + "node": "conv1d_4", + "inputs": [ + "cat_9", + "p_model_model_language_model_layers_5_linear_attn_conv1d_weight" + ], + "output": "conv1d_4", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_53", + "inputs": [ + "conv1d_4" + ], + "output": "slice_53", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_13", + "inputs": [ + "slice_53" + ], + "output": "silu_13", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_54", + "inputs": [ + "silu_13" + ], + "output": "slice_54", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_31", + "inputs": [ + "slice_54" + ], + "output": "transpose_31", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_55", + "inputs": [ + "transpose_31" + ], + "output": "slice_55", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_56", + "inputs": [ + "transpose_31" + ], + "output": "slice_56", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_57", + "inputs": [ + "transpose_31" + ], + "output": "slice_57", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_5", + "inputs": [ + "linear_41" + ], + "output": "sigmoid_5", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_4", + "inputs": [ + "p_model_model_language_model_layers_5_linear_attn_a_log" + ], + "output": "exp_4", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_6", + "inputs": [ + "exp_4" + ], + "output": "neg_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_55", + "inputs": [ + "linear_42", + "p_model_model_language_model_layers_5_linear_attn_dt_bias" + ], + "output": "add_55", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_4", + "inputs": [ + "add_55" + ], + "output": "softplus_4", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_71", + "inputs": [ + "neg_6", + "softplus_4" + ], + "output": "mul_71", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_72", + "inputs": [ + "slice_55" + ], + "output": "mul_72", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_9", + "inputs": [ + "mul_72" + ], + "output": "sum_9", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_56", + "inputs": [ + "sum_9" + ], + "output": "add_56", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_25", + "inputs": [ + "add_56" + ], + "output": "rsqrt_25", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_73", + "inputs": [ + "slice_55", + "rsqrt_25" + ], + "output": "mul_73", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_74", + "inputs": [ + "slice_56" + ], + "output": "mul_74", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_10", + "inputs": [ + "mul_74" + ], + "output": "sum_10", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_57", + "inputs": [ + "sum_10" + ], + "output": "add_57", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_26", + "inputs": [ + "add_57" + ], + "output": "rsqrt_26", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_75", + "inputs": [ + "slice_56", + "rsqrt_26" + ], + "output": "mul_75", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_4", + "inputs": [ + "mul_73", + "mul_75", + "slice_57", + "mul_71", + "sigmoid_5", + "states_11" + ], + "output": "normalized_gated_delta_rule_4", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_58", + "inputs": [ + "normalized_gated_delta_rule_4" + ], + "output": "slice_58", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_11" + }, + { + "node": "slice_59", + "inputs": [ + "normalized_gated_delta_rule_4" + ], + "output": "slice_59", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_32", + "inputs": [ + "slice_59" + ], + "output": "transpose_32", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_77", + "inputs": [ + "transpose_32", + "p_model_model_language_model_layers_5_linear_attn_norm_weight" + ], + "output": "mul_77", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_78", + "inputs": [ + "linear_40", + "mul_77" + ], + "output": "mul_78", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_43", + "inputs": [ + "mul_78", + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_43", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_59", + "inputs": [ + "add_52", + "linear_43" + ], + "output": "add_59", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_80", + "inputs": [ + "add_59", + "p_model_model_language_model_layers_5_post_attention_layernorm_weight" + ], + "output": "mul_80", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_44", + "inputs": [ + "mul_80", + "p_model_model_language_model_layers_5_mlp_gate_proj_weight", + "p_model_model_language_model_layers_5_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_44", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_45", + "inputs": [ + "mul_80", + "p_model_model_language_model_layers_5_mlp_up_proj_weight", + "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_45", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_81", + "inputs": [ + "linear_44", + "linear_45" + ], + "output": "mul_81", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_46", + "inputs": [ + "mul_81", + "p_model_model_language_model_layers_5_mlp_down_proj_weight", + "p_model_model_language_model_layers_5_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_46", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_62", + "inputs": [ + "add_59", + "linear_46" + ], + "output": "add_62", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_83", + "inputs": [ + "add_62", + "p_model_model_language_model_layers_6_input_layernorm_weight" + ], + "output": "mul_83", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_47", + "inputs": [ + "mul_83", + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_47", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_33", + "inputs": [ + "linear_47" + ], + "output": "transpose_33", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_48", + "inputs": [ + "mul_83", + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_48", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_49", + "inputs": [ + "mul_83", + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_49", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_50", + "inputs": [ + "mul_83", + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_50", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_10", + "inputs": [ + "states_12", + "transpose_33" + ], + "output": "cat_10", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_60", + "inputs": [ + "cat_10" + ], + "output": "slice_60", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_12" + }, + { + "node": "conv1d_5", + "inputs": [ + "cat_10", + "p_model_model_language_model_layers_6_linear_attn_conv1d_weight" + ], + "output": "conv1d_5", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_61", + "inputs": [ + "conv1d_5" + ], + "output": "slice_61", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_16", + "inputs": [ + "slice_61" + ], + "output": "silu_16", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_62", + "inputs": [ + "silu_16" + ], + "output": "slice_62", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_34", + "inputs": [ + "slice_62" + ], + "output": "transpose_34", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_63", + "inputs": [ + "transpose_34" + ], + "output": "slice_63", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_64", + "inputs": [ + "transpose_34" + ], + "output": "slice_64", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_65", + "inputs": [ + "transpose_34" + ], + "output": "slice_65", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_6", + "inputs": [ + "linear_49" + ], + "output": "sigmoid_6", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_5", + "inputs": [ + "p_model_model_language_model_layers_6_linear_attn_a_log" + ], + "output": "exp_5", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_7", + "inputs": [ + "exp_5" + ], + "output": "neg_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_65", + "inputs": [ + "linear_50", + "p_model_model_language_model_layers_6_linear_attn_dt_bias" + ], + "output": "add_65", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_5", + "inputs": [ + "add_65" + ], + "output": "softplus_5", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_84", + "inputs": [ + "neg_7", + "softplus_5" + ], + "output": "mul_84", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_85", + "inputs": [ + "slice_63" + ], + "output": "mul_85", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_11", + "inputs": [ + "mul_85" + ], + "output": "sum_11", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_66", + "inputs": [ + "sum_11" + ], + "output": "add_66", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_30", + "inputs": [ + "add_66" + ], + "output": "rsqrt_30", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_86", + "inputs": [ + "slice_63", + "rsqrt_30" + ], + "output": "mul_86", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_87", + "inputs": [ + "slice_64" + ], + "output": "mul_87", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_12", + "inputs": [ + "mul_87" + ], + "output": "sum_12", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_67", + "inputs": [ + "sum_12" + ], + "output": "add_67", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_31", + "inputs": [ + "add_67" + ], + "output": "rsqrt_31", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_88", + "inputs": [ + "slice_64", + "rsqrt_31" + ], + "output": "mul_88", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_5", + "inputs": [ + "mul_86", + "mul_88", + "slice_65", + "mul_84", + "sigmoid_6", + "states_13" + ], + "output": "normalized_gated_delta_rule_5", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_66", + "inputs": [ + "normalized_gated_delta_rule_5" + ], + "output": "slice_66", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_13" + }, + { + "node": "slice_67", + "inputs": [ + "normalized_gated_delta_rule_5" + ], + "output": "slice_67", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_35", + "inputs": [ + "slice_67" + ], + "output": "transpose_35", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_90", + "inputs": [ + "transpose_35", + "p_model_model_language_model_layers_6_linear_attn_norm_weight" + ], + "output": "mul_90", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_91", + "inputs": [ + "linear_48", + "mul_90" + ], + "output": "mul_91", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_51", + "inputs": [ + "mul_91", + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_51", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_69", + "inputs": [ + "add_62", + "linear_51" + ], + "output": "add_69", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_93", + "inputs": [ + "add_69", + "p_model_model_language_model_layers_6_post_attention_layernorm_weight" + ], + "output": "mul_93", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_52", + "inputs": [ + "mul_93", + "p_model_model_language_model_layers_6_mlp_gate_proj_weight", + "p_model_model_language_model_layers_6_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_52", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_53", + "inputs": [ + "mul_93", + "p_model_model_language_model_layers_6_mlp_up_proj_weight", + "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_53", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_94", + "inputs": [ + "linear_52", + "linear_53" + ], + "output": "mul_94", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_54", + "inputs": [ + "mul_94", + "p_model_model_language_model_layers_6_mlp_down_proj_weight", + "p_model_model_language_model_layers_6_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_54", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_72", + "inputs": [ + "add_69", + "linear_54" + ], + "output": "add_72", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_96", + "inputs": [ + "add_72", + "p_model_model_language_model_layers_7_input_layernorm_weight" + ], + "output": "mul_96", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_55", + "inputs": [ + "mul_96", + "p_model_model_language_model_layers_7_self_attn_q_proj_weight", + "p_model_model_language_model_layers_7_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_55", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d1e9766c981b6c437331275b2afdfc74e0789005b52c2015839fc2394961da1b.wgsl" + }, + { + "node": "slice_68", + "inputs": [ + "linear_55" + ], + "output": "slice_68", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3a76085983a5330bf07a4166ea4975700403f9d8035a8ef450f7f18b9cf47c66.wgsl" + }, + { + "node": "slice_69", + "inputs": [ + "linear_55" + ], + "output": "slice_69", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9044193f1f4724f15ef5fb41040b7fbac5f62fcaea80594737214bf9cd6be41a.wgsl" + }, + { + "node": "mul_98", + "inputs": [ + "slice_68", + "p_model_model_language_model_layers_7_self_attn_q_norm_weight" + ], + "output": "mul_98", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/ef364f47cced448917d1b6842c708dee101d53d7a781014f9a822214cea72bf8.wgsl" + }, + { + "node": "transpose_36", + "inputs": [ + "mul_98" + ], + "output": "transpose_36", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/1d0cbb068fe17123e646934bfa1259f19e5db04e45d72a157d3ea4fa48c81aa0.wgsl" + }, + { + "node": "linear_56", + "inputs": [ + "mul_96", + "p_model_model_language_model_layers_7_self_attn_k_proj_weight", + "p_model_model_language_model_layers_7_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_56", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/c5b7564d87c89606da1cba30e05e401e4fb8c83957e779a694d2ec2c472d88ad.wgsl" + }, + { + "node": "mul_100", + "inputs": [ + "linear_56", + "p_model_model_language_model_layers_7_self_attn_k_norm_weight" + ], + "output": "mul_100", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl" + }, + { + "node": "transpose_37", + "inputs": [ + "mul_100" + ], + "output": "transpose_37", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/bcb78b92dbbfeb212c190e064bc2f9d8e79c201398cf38c9aa62df6fdba94469.wgsl" + }, + { + "node": "linear_57", + "inputs": [ + "mul_96", + "p_model_model_language_model_layers_7_self_attn_v_proj_weight", + "p_model_model_language_model_layers_7_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_57", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/c5b7564d87c89606da1cba30e05e401e4fb8c83957e779a694d2ec2c472d88ad.wgsl" + }, + { + "node": "transpose_38", + "inputs": [ + "linear_57" + ], + "output": "transpose_38", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/bcb78b92dbbfeb212c190e064bc2f9d8e79c201398cf38c9aa62df6fdba94469.wgsl" + }, + { + "node": "slice_70", + "inputs": [ + "transpose_36" + ], + "output": "slice_70", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/4d8ebffbbf193b18c13a5cc17d286bf5c92c5ba53ca022af78a9151ec1e5b9c0.wgsl" + }, + { + "node": "slice_71", + "inputs": [ + "transpose_36" + ], + "output": "slice_71", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/4406ab1e6e4b9eeb91ace297320aa05a55cc481e02894a4a74b655b4070ade36.wgsl" + }, + { + "node": "slice_72", + "inputs": [ + "transpose_37" + ], + "output": "slice_72", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/89f98b79c18e7f4bbbd79cbd8d3e80108d0e1e4407c161d9637b6a88d3fbe06b.wgsl" + }, + { + "node": "slice_73", + "inputs": [ + "transpose_37" + ], + "output": "slice_73", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd70e4fc42ac55f0f52612841bcc3f9fe7df5bbb0fc76f94e56a131a92454658.wgsl" + }, + { + "node": "mul_101", + "inputs": [ + "slice_70", + "mul" + ], + "output": "mul_101", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl" + }, + { + "node": "slice_74", + "inputs": [ + "slice_70" + ], + "output": "slice_74", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/2c66dcac8fdd3caa9ff1e2441524070c21c24df9526bee44c49e765a8d4c8222.wgsl" + }, + { + "node": "slice_75", + "inputs": [ + "slice_70" + ], + "output": "slice_75", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/33e8ec04c66adb4988f562f0f9d3e92d29dfda91a56d0ed62be598c7c737f5ba.wgsl" + }, + { + "node": "neg_8", + "inputs": [ + "slice_75" + ], + "output": "neg_8", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/3e34409407100d4858c4c109cf62f3eaf2f4c6081166b0ec8679105e6dfbe7fe.wgsl" + }, + { + "node": "cat_11", + "inputs": [ + "neg_8", + "slice_74" + ], + "output": "cat_11", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/85964f8bbe5a2e4d430930e1050ea76871f0f2453c3cd4ba020002c0562ca680.wgsl" + }, + { + "node": "mul_102", + "inputs": [ + "cat_11", + "mul_1" + ], + "output": "mul_102", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl" + }, + { + "node": "add_79", + "inputs": [ + "mul_101", + "mul_102" + ], + "output": "add_79", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/d2866fd21f58839ea7d43bf83fde4243c38e4f1771034520abb020be5d819a96.wgsl" + }, + { + "node": "mul_103", + "inputs": [ + "slice_72", + "mul" + ], + "output": "mul_103", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl" + }, + { + "node": "slice_76", + "inputs": [ + "slice_72" + ], + "output": "slice_76", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/7b7c42c04ea3c6695ddead641e88b82d8f130015e138fc323eec509b5b56cfa9.wgsl" + }, + { + "node": "slice_77", + "inputs": [ + "slice_72" + ], + "output": "slice_77", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/6d82b7ed99a04f03fa0b9d241188d9613fa3ce9ff425cf8ee30a7f4ce6ded13d.wgsl" + }, + { + "node": "neg_9", + "inputs": [ + "slice_77" + ], + "output": "neg_9", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/adf1e58a124e1301625024ae142dca5fb1dd50799fa94b52eaeffb82ea99852c.wgsl" + }, + { + "node": "cat_12", + "inputs": [ + "neg_9", + "slice_76" + ], + "output": "cat_12", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl" + }, + { + "node": "mul_104", + "inputs": [ + "cat_12", + "mul_1" + ], + "output": "mul_104", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl" + }, + { + "node": "add_80", + "inputs": [ + "mul_103", + "mul_104" + ], + "output": "add_80", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl" + }, + { + "node": "cat_13", + "inputs": [ + "add_79", + "slice_71" + ], + "output": "cat_13", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/8affc7c0e85127430e406be35ebb5c5354a8dd26ecc4ae9ab6ace80ce6775f46.wgsl" + }, + { + "node": "cat_14", + "inputs": [ + "add_80", + "slice_73" + ], + "output": "cat_14", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl" + }, + { + "node": "cache_update_2", + "inputs": [ + "cat_14", + "position_ids" + ], + "output": "cache_update_2", + "workgroups": [ + 512, + 1, + 1 + ], + "inPlaceOf": "states_14", + "kernel": "kernels/e369b5c120d84bfebbc35e3da096c33e55d20a607aecd5d82da5de4c95176e00.wgsl" + }, + { + "node": "cache_update_3", + "inputs": [ + "transpose_38", + "position_ids" + ], + "output": "cache_update_3", + "workgroups": [ + 512, + 1, + 1 + ], + "inPlaceOf": "states_15", + "kernel": "kernels/e369b5c120d84bfebbc35e3da096c33e55d20a607aecd5d82da5de4c95176e00.wgsl" + }, + { + "node": "matmul_4", + "inputs": [ + "cat_13", + "cache_update_2", + "cache_update_3", + "attention_mask" + ], + "output": "matmul_4", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/140ebe12d31f1420a8a159b91ae1773390190f496e06e0e5d7b5351f38245068.wgsl" + }, + { + "node": "transpose_40", + "inputs": [ + "matmul_4" + ], + "output": "transpose_40", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4f2c0d08150d4b97d90c244db2d41baca81484c2630dbbd6d586f2d8a4b07b8b.wgsl" + }, + { + "node": "sigmoid_7", + "inputs": [ + "slice_69" + ], + "output": "sigmoid_7", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/50535e4c99d9fb4a272d8260a1415f6b38a7ad1025da64e145f230df697d9b73.wgsl" + }, + { + "node": "mul_106", + "inputs": [ + "transpose_40", + "sigmoid_7" + ], + "output": "mul_106", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/a9f6bc361f1ea31f79f0ef6164872eeb04896a45338871a2d779789b47b1920b.wgsl" + }, + { + "node": "linear_58", + "inputs": [ + "mul_106", + "p_model_model_language_model_layers_7_self_attn_o_proj_weight", + "p_model_model_language_model_layers_7_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_58", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_82", + "inputs": [ + "add_72", + "linear_58" + ], + "output": "add_82", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_108", + "inputs": [ + "add_82", + "p_model_model_language_model_layers_7_post_attention_layernorm_weight" + ], + "output": "mul_108", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_59", + "inputs": [ + "mul_108", + "p_model_model_language_model_layers_7_mlp_gate_proj_weight", + "p_model_model_language_model_layers_7_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_59", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_60", + "inputs": [ + "mul_108", + "p_model_model_language_model_layers_7_mlp_up_proj_weight", + "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_60", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_109", + "inputs": [ + "linear_59", + "linear_60" + ], + "output": "mul_109", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_61", + "inputs": [ + "mul_109", + "p_model_model_language_model_layers_7_mlp_down_proj_weight", + "p_model_model_language_model_layers_7_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_61", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_85", + "inputs": [ + "add_82", + "linear_61" + ], + "output": "add_85", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_111", + "inputs": [ + "add_85", + "p_model_model_language_model_layers_8_input_layernorm_weight" + ], + "output": "mul_111", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_62", + "inputs": [ + "mul_111", + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_62", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_41", + "inputs": [ + "linear_62" + ], + "output": "transpose_41", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_63", + "inputs": [ + "mul_111", + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_63", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_64", + "inputs": [ + "mul_111", + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_64", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_65", + "inputs": [ + "mul_111", + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_65", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_15", + "inputs": [ + "states_16", + "transpose_41" + ], + "output": "cat_15", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_78", + "inputs": [ + "cat_15" + ], + "output": "slice_78", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_16" + }, + { + "node": "conv1d_6", + "inputs": [ + "cat_15", + "p_model_model_language_model_layers_8_linear_attn_conv1d_weight" + ], + "output": "conv1d_6", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_79", + "inputs": [ + "conv1d_6" + ], + "output": "slice_79", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_20", + "inputs": [ + "slice_79" + ], + "output": "silu_20", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_80", + "inputs": [ + "silu_20" + ], + "output": "slice_80", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_42", + "inputs": [ + "slice_80" + ], + "output": "transpose_42", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_81", + "inputs": [ + "transpose_42" + ], + "output": "slice_81", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_82", + "inputs": [ + "transpose_42" + ], + "output": "slice_82", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_83", + "inputs": [ + "transpose_42" + ], + "output": "slice_83", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_8", + "inputs": [ + "linear_64" + ], + "output": "sigmoid_8", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_6", + "inputs": [ + "p_model_model_language_model_layers_8_linear_attn_a_log" + ], + "output": "exp_6", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_10", + "inputs": [ + "exp_6" + ], + "output": "neg_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_88", + "inputs": [ + "linear_65", + "p_model_model_language_model_layers_8_linear_attn_dt_bias" + ], + "output": "add_88", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_6", + "inputs": [ + "add_88" + ], + "output": "softplus_6", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_112", + "inputs": [ + "neg_10", + "softplus_6" + ], + "output": "mul_112", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_113", + "inputs": [ + "slice_81" + ], + "output": "mul_113", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_13", + "inputs": [ + "mul_113" + ], + "output": "sum_13", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_89", + "inputs": [ + "sum_13" + ], + "output": "add_89", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_39", + "inputs": [ + "add_89" + ], + "output": "rsqrt_39", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_114", + "inputs": [ + "slice_81", + "rsqrt_39" + ], + "output": "mul_114", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_115", + "inputs": [ + "slice_82" + ], + "output": "mul_115", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_14", + "inputs": [ + "mul_115" + ], + "output": "sum_14", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_90", + "inputs": [ + "sum_14" + ], + "output": "add_90", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_40", + "inputs": [ + "add_90" + ], + "output": "rsqrt_40", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_116", + "inputs": [ + "slice_82", + "rsqrt_40" + ], + "output": "mul_116", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_6", + "inputs": [ + "mul_114", + "mul_116", + "slice_83", + "mul_112", + "sigmoid_8", + "states_17" + ], + "output": "normalized_gated_delta_rule_6", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_84", + "inputs": [ + "normalized_gated_delta_rule_6" + ], + "output": "slice_84", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_17" + }, + { + "node": "slice_85", + "inputs": [ + "normalized_gated_delta_rule_6" + ], + "output": "slice_85", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_43", + "inputs": [ + "slice_85" + ], + "output": "transpose_43", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_118", + "inputs": [ + "transpose_43", + "p_model_model_language_model_layers_8_linear_attn_norm_weight" + ], + "output": "mul_118", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_119", + "inputs": [ + "linear_63", + "mul_118" + ], + "output": "mul_119", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_66", + "inputs": [ + "mul_119", + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_66", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_92", + "inputs": [ + "add_85", + "linear_66" + ], + "output": "add_92", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_121", + "inputs": [ + "add_92", + "p_model_model_language_model_layers_8_post_attention_layernorm_weight" + ], + "output": "mul_121", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_67", + "inputs": [ + "mul_121", + "p_model_model_language_model_layers_8_mlp_gate_proj_weight", + "p_model_model_language_model_layers_8_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_67", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_68", + "inputs": [ + "mul_121", + "p_model_model_language_model_layers_8_mlp_up_proj_weight", + "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_68", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_122", + "inputs": [ + "linear_67", + "linear_68" + ], + "output": "mul_122", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_69", + "inputs": [ + "mul_122", + "p_model_model_language_model_layers_8_mlp_down_proj_weight", + "p_model_model_language_model_layers_8_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_69", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_95", + "inputs": [ + "add_92", + "linear_69" + ], + "output": "add_95", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_124", + "inputs": [ + "add_95", + "p_model_model_language_model_layers_9_input_layernorm_weight" + ], + "output": "mul_124", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_70", + "inputs": [ + "mul_124", + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_70", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_44", + "inputs": [ + "linear_70" + ], + "output": "transpose_44", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_71", + "inputs": [ + "mul_124", + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_71", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_72", + "inputs": [ + "mul_124", + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_72", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_73", + "inputs": [ + "mul_124", + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_73", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_16", + "inputs": [ + "states_18", + "transpose_44" + ], + "output": "cat_16", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_86", + "inputs": [ + "cat_16" + ], + "output": "slice_86", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_18" + }, + { + "node": "conv1d_7", + "inputs": [ + "cat_16", + "p_model_model_language_model_layers_9_linear_attn_conv1d_weight" + ], + "output": "conv1d_7", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_87", + "inputs": [ + "conv1d_7" + ], + "output": "slice_87", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_23", + "inputs": [ + "slice_87" + ], + "output": "silu_23", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_88", + "inputs": [ + "silu_23" + ], + "output": "slice_88", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_45", + "inputs": [ + "slice_88" + ], + "output": "transpose_45", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_89", + "inputs": [ + "transpose_45" + ], + "output": "slice_89", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_90", + "inputs": [ + "transpose_45" + ], + "output": "slice_90", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_91", + "inputs": [ + "transpose_45" + ], + "output": "slice_91", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_9", + "inputs": [ + "linear_72" + ], + "output": "sigmoid_9", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_7", + "inputs": [ + "p_model_model_language_model_layers_9_linear_attn_a_log" + ], + "output": "exp_7", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_11", + "inputs": [ + "exp_7" + ], + "output": "neg_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_98", + "inputs": [ + "linear_73", + "p_model_model_language_model_layers_9_linear_attn_dt_bias" + ], + "output": "add_98", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_7", + "inputs": [ + "add_98" + ], + "output": "softplus_7", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_125", + "inputs": [ + "neg_11", + "softplus_7" + ], + "output": "mul_125", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_126", + "inputs": [ + "slice_89" + ], + "output": "mul_126", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_15", + "inputs": [ + "mul_126" + ], + "output": "sum_15", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_99", + "inputs": [ + "sum_15" + ], + "output": "add_99", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_44", + "inputs": [ + "add_99" + ], + "output": "rsqrt_44", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_127", + "inputs": [ + "slice_89", + "rsqrt_44" + ], + "output": "mul_127", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_128", + "inputs": [ + "slice_90" + ], + "output": "mul_128", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_16", + "inputs": [ + "mul_128" + ], + "output": "sum_16", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_100", + "inputs": [ + "sum_16" + ], + "output": "add_100", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_45", + "inputs": [ + "add_100" + ], + "output": "rsqrt_45", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_129", + "inputs": [ + "slice_90", + "rsqrt_45" + ], + "output": "mul_129", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_7", + "inputs": [ + "mul_127", + "mul_129", + "slice_91", + "mul_125", + "sigmoid_9", + "states_19" + ], + "output": "normalized_gated_delta_rule_7", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_92", + "inputs": [ + "normalized_gated_delta_rule_7" + ], + "output": "slice_92", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_19" + }, + { + "node": "slice_93", + "inputs": [ + "normalized_gated_delta_rule_7" + ], + "output": "slice_93", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_46", + "inputs": [ + "slice_93" + ], + "output": "transpose_46", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_131", + "inputs": [ + "transpose_46", + "p_model_model_language_model_layers_9_linear_attn_norm_weight" + ], + "output": "mul_131", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_132", + "inputs": [ + "linear_71", + "mul_131" + ], + "output": "mul_132", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_74", + "inputs": [ + "mul_132", + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_74", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_102", + "inputs": [ + "add_95", + "linear_74" + ], + "output": "add_102", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_134", + "inputs": [ + "add_102", + "p_model_model_language_model_layers_9_post_attention_layernorm_weight" + ], + "output": "mul_134", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_75", + "inputs": [ + "mul_134", + "p_model_model_language_model_layers_9_mlp_gate_proj_weight", + "p_model_model_language_model_layers_9_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_75", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_76", + "inputs": [ + "mul_134", + "p_model_model_language_model_layers_9_mlp_up_proj_weight", + "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_76", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_135", + "inputs": [ + "linear_75", + "linear_76" + ], + "output": "mul_135", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_77", + "inputs": [ + "mul_135", + "p_model_model_language_model_layers_9_mlp_down_proj_weight", + "p_model_model_language_model_layers_9_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_77", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_105", + "inputs": [ + "add_102", + "linear_77" + ], + "output": "add_105", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_137", + "inputs": [ + "add_105", + "p_model_model_language_model_layers_10_input_layernorm_weight" + ], + "output": "mul_137", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_78", + "inputs": [ + "mul_137", + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_78", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_47", + "inputs": [ + "linear_78" + ], + "output": "transpose_47", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_79", + "inputs": [ + "mul_137", + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_79", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_80", + "inputs": [ + "mul_137", + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_80", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_81", + "inputs": [ + "mul_137", + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_81", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_17", + "inputs": [ + "states_20", + "transpose_47" + ], + "output": "cat_17", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_94", + "inputs": [ + "cat_17" + ], + "output": "slice_94", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_20" + }, + { + "node": "conv1d_8", + "inputs": [ + "cat_17", + "p_model_model_language_model_layers_10_linear_attn_conv1d_weight" + ], + "output": "conv1d_8", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_95", + "inputs": [ + "conv1d_8" + ], + "output": "slice_95", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_26", + "inputs": [ + "slice_95" + ], + "output": "silu_26", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_96", + "inputs": [ + "silu_26" + ], + "output": "slice_96", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_48", + "inputs": [ + "slice_96" + ], + "output": "transpose_48", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_97", + "inputs": [ + "transpose_48" + ], + "output": "slice_97", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_98", + "inputs": [ + "transpose_48" + ], + "output": "slice_98", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_99", + "inputs": [ + "transpose_48" + ], + "output": "slice_99", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_10", + "inputs": [ + "linear_80" + ], + "output": "sigmoid_10", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_8", + "inputs": [ + "p_model_model_language_model_layers_10_linear_attn_a_log" + ], + "output": "exp_8", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_12", + "inputs": [ + "exp_8" + ], + "output": "neg_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_108", + "inputs": [ + "linear_81", + "p_model_model_language_model_layers_10_linear_attn_dt_bias" + ], + "output": "add_108", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_8", + "inputs": [ + "add_108" + ], + "output": "softplus_8", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_138", + "inputs": [ + "neg_12", + "softplus_8" + ], + "output": "mul_138", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_139", + "inputs": [ + "slice_97" + ], + "output": "mul_139", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_17", + "inputs": [ + "mul_139" + ], + "output": "sum_17", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_109", + "inputs": [ + "sum_17" + ], + "output": "add_109", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_49", + "inputs": [ + "add_109" + ], + "output": "rsqrt_49", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_140", + "inputs": [ + "slice_97", + "rsqrt_49" + ], + "output": "mul_140", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_141", + "inputs": [ + "slice_98" + ], + "output": "mul_141", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_18", + "inputs": [ + "mul_141" + ], + "output": "sum_18", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_110", + "inputs": [ + "sum_18" + ], + "output": "add_110", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_50", + "inputs": [ + "add_110" + ], + "output": "rsqrt_50", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_142", + "inputs": [ + "slice_98", + "rsqrt_50" + ], + "output": "mul_142", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_8", + "inputs": [ + "mul_140", + "mul_142", + "slice_99", + "mul_138", + "sigmoid_10", + "states_21" + ], + "output": "normalized_gated_delta_rule_8", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_100", + "inputs": [ + "normalized_gated_delta_rule_8" + ], + "output": "slice_100", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_21" + }, + { + "node": "slice_101", + "inputs": [ + "normalized_gated_delta_rule_8" + ], + "output": "slice_101", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_49", + "inputs": [ + "slice_101" + ], + "output": "transpose_49", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_144", + "inputs": [ + "transpose_49", + "p_model_model_language_model_layers_10_linear_attn_norm_weight" + ], + "output": "mul_144", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_145", + "inputs": [ + "linear_79", + "mul_144" + ], + "output": "mul_145", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_82", + "inputs": [ + "mul_145", + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_82", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_112", + "inputs": [ + "add_105", + "linear_82" + ], + "output": "add_112", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_147", + "inputs": [ + "add_112", + "p_model_model_language_model_layers_10_post_attention_layernorm_weight" + ], + "output": "mul_147", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_83", + "inputs": [ + "mul_147", + "p_model_model_language_model_layers_10_mlp_gate_proj_weight", + "p_model_model_language_model_layers_10_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_83", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_84", + "inputs": [ + "mul_147", + "p_model_model_language_model_layers_10_mlp_up_proj_weight", + "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_84", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_148", + "inputs": [ + "linear_83", + "linear_84" + ], + "output": "mul_148", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_85", + "inputs": [ + "mul_148", + "p_model_model_language_model_layers_10_mlp_down_proj_weight", + "p_model_model_language_model_layers_10_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_85", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_115", + "inputs": [ + "add_112", + "linear_85" + ], + "output": "add_115", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_150", + "inputs": [ + "add_115", + "p_model_model_language_model_layers_11_input_layernorm_weight" + ], + "output": "mul_150", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_86", + "inputs": [ + "mul_150", + "p_model_model_language_model_layers_11_self_attn_q_proj_weight", + "p_model_model_language_model_layers_11_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_86", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d1e9766c981b6c437331275b2afdfc74e0789005b52c2015839fc2394961da1b.wgsl" + }, + { + "node": "slice_102", + "inputs": [ + "linear_86" + ], + "output": "slice_102", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3a76085983a5330bf07a4166ea4975700403f9d8035a8ef450f7f18b9cf47c66.wgsl" + }, + { + "node": "slice_103", + "inputs": [ + "linear_86" + ], + "output": "slice_103", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9044193f1f4724f15ef5fb41040b7fbac5f62fcaea80594737214bf9cd6be41a.wgsl" + }, + { + "node": "mul_152", + "inputs": [ + "slice_102", + "p_model_model_language_model_layers_11_self_attn_q_norm_weight" + ], + "output": "mul_152", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/ef364f47cced448917d1b6842c708dee101d53d7a781014f9a822214cea72bf8.wgsl" + }, + { + "node": "transpose_50", + "inputs": [ + "mul_152" + ], + "output": "transpose_50", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/1d0cbb068fe17123e646934bfa1259f19e5db04e45d72a157d3ea4fa48c81aa0.wgsl" + }, + { + "node": "linear_87", + "inputs": [ + "mul_150", + "p_model_model_language_model_layers_11_self_attn_k_proj_weight", + "p_model_model_language_model_layers_11_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_87", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/c5b7564d87c89606da1cba30e05e401e4fb8c83957e779a694d2ec2c472d88ad.wgsl" + }, + { + "node": "mul_154", + "inputs": [ + "linear_87", + "p_model_model_language_model_layers_11_self_attn_k_norm_weight" + ], + "output": "mul_154", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl" + }, + { + "node": "transpose_51", + "inputs": [ + "mul_154" + ], + "output": "transpose_51", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/bcb78b92dbbfeb212c190e064bc2f9d8e79c201398cf38c9aa62df6fdba94469.wgsl" + }, + { + "node": "linear_88", + "inputs": [ + "mul_150", + "p_model_model_language_model_layers_11_self_attn_v_proj_weight", + "p_model_model_language_model_layers_11_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_88", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/c5b7564d87c89606da1cba30e05e401e4fb8c83957e779a694d2ec2c472d88ad.wgsl" + }, + { + "node": "transpose_52", + "inputs": [ + "linear_88" + ], + "output": "transpose_52", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/bcb78b92dbbfeb212c190e064bc2f9d8e79c201398cf38c9aa62df6fdba94469.wgsl" + }, + { + "node": "slice_104", + "inputs": [ + "transpose_50" + ], + "output": "slice_104", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/4d8ebffbbf193b18c13a5cc17d286bf5c92c5ba53ca022af78a9151ec1e5b9c0.wgsl" + }, + { + "node": "slice_105", + "inputs": [ + "transpose_50" + ], + "output": "slice_105", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/4406ab1e6e4b9eeb91ace297320aa05a55cc481e02894a4a74b655b4070ade36.wgsl" + }, + { + "node": "slice_106", + "inputs": [ + "transpose_51" + ], + "output": "slice_106", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/89f98b79c18e7f4bbbd79cbd8d3e80108d0e1e4407c161d9637b6a88d3fbe06b.wgsl" + }, + { + "node": "slice_107", + "inputs": [ + "transpose_51" + ], + "output": "slice_107", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd70e4fc42ac55f0f52612841bcc3f9fe7df5bbb0fc76f94e56a131a92454658.wgsl" + }, + { + "node": "mul_155", + "inputs": [ + "slice_104", + "mul" + ], + "output": "mul_155", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl" + }, + { + "node": "slice_108", + "inputs": [ + "slice_104" + ], + "output": "slice_108", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/2c66dcac8fdd3caa9ff1e2441524070c21c24df9526bee44c49e765a8d4c8222.wgsl" + }, + { + "node": "slice_109", + "inputs": [ + "slice_104" + ], + "output": "slice_109", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/33e8ec04c66adb4988f562f0f9d3e92d29dfda91a56d0ed62be598c7c737f5ba.wgsl" + }, + { + "node": "neg_13", + "inputs": [ + "slice_109" + ], + "output": "neg_13", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/3e34409407100d4858c4c109cf62f3eaf2f4c6081166b0ec8679105e6dfbe7fe.wgsl" + }, + { + "node": "cat_18", + "inputs": [ + "neg_13", + "slice_108" + ], + "output": "cat_18", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/85964f8bbe5a2e4d430930e1050ea76871f0f2453c3cd4ba020002c0562ca680.wgsl" + }, + { + "node": "mul_156", + "inputs": [ + "cat_18", + "mul_1" + ], + "output": "mul_156", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl" + }, + { + "node": "add_122", + "inputs": [ + "mul_155", + "mul_156" + ], + "output": "add_122", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/d2866fd21f58839ea7d43bf83fde4243c38e4f1771034520abb020be5d819a96.wgsl" + }, + { + "node": "mul_157", + "inputs": [ + "slice_106", + "mul" + ], + "output": "mul_157", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl" + }, + { + "node": "slice_110", + "inputs": [ + "slice_106" + ], + "output": "slice_110", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/7b7c42c04ea3c6695ddead641e88b82d8f130015e138fc323eec509b5b56cfa9.wgsl" + }, + { + "node": "slice_111", + "inputs": [ + "slice_106" + ], + "output": "slice_111", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/6d82b7ed99a04f03fa0b9d241188d9613fa3ce9ff425cf8ee30a7f4ce6ded13d.wgsl" + }, + { + "node": "neg_14", + "inputs": [ + "slice_111" + ], + "output": "neg_14", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/adf1e58a124e1301625024ae142dca5fb1dd50799fa94b52eaeffb82ea99852c.wgsl" + }, + { + "node": "cat_19", + "inputs": [ + "neg_14", + "slice_110" + ], + "output": "cat_19", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl" + }, + { + "node": "mul_158", + "inputs": [ + "cat_19", + "mul_1" + ], + "output": "mul_158", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl" + }, + { + "node": "add_123", + "inputs": [ + "mul_157", + "mul_158" + ], + "output": "add_123", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl" + }, + { + "node": "cat_20", + "inputs": [ + "add_122", + "slice_105" + ], + "output": "cat_20", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/8affc7c0e85127430e406be35ebb5c5354a8dd26ecc4ae9ab6ace80ce6775f46.wgsl" + }, + { + "node": "cat_21", + "inputs": [ + "add_123", + "slice_107" + ], + "output": "cat_21", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl" + }, + { + "node": "cache_update_4", + "inputs": [ + "cat_21", + "position_ids" + ], + "output": "cache_update_4", + "workgroups": [ + 512, + 1, + 1 + ], + "inPlaceOf": "states_22", + "kernel": "kernels/e369b5c120d84bfebbc35e3da096c33e55d20a607aecd5d82da5de4c95176e00.wgsl" + }, + { + "node": "cache_update_5", + "inputs": [ + "transpose_52", + "position_ids" + ], + "output": "cache_update_5", + "workgroups": [ + 512, + 1, + 1 + ], + "inPlaceOf": "states_23", + "kernel": "kernels/e369b5c120d84bfebbc35e3da096c33e55d20a607aecd5d82da5de4c95176e00.wgsl" + }, + { + "node": "matmul_6", + "inputs": [ + "cat_20", + "cache_update_4", + "cache_update_5", + "attention_mask" + ], + "output": "matmul_6", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/140ebe12d31f1420a8a159b91ae1773390190f496e06e0e5d7b5351f38245068.wgsl" + }, + { + "node": "transpose_54", + "inputs": [ + "matmul_6" + ], + "output": "transpose_54", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4f2c0d08150d4b97d90c244db2d41baca81484c2630dbbd6d586f2d8a4b07b8b.wgsl" + }, + { + "node": "sigmoid_11", + "inputs": [ + "slice_103" + ], + "output": "sigmoid_11", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/50535e4c99d9fb4a272d8260a1415f6b38a7ad1025da64e145f230df697d9b73.wgsl" + }, + { + "node": "mul_160", + "inputs": [ + "transpose_54", + "sigmoid_11" + ], + "output": "mul_160", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/a9f6bc361f1ea31f79f0ef6164872eeb04896a45338871a2d779789b47b1920b.wgsl" + }, + { + "node": "linear_89", + "inputs": [ + "mul_160", + "p_model_model_language_model_layers_11_self_attn_o_proj_weight", + "p_model_model_language_model_layers_11_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_89", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_125", + "inputs": [ + "add_115", + "linear_89" + ], + "output": "add_125", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_162", + "inputs": [ + "add_125", + "p_model_model_language_model_layers_11_post_attention_layernorm_weight" + ], + "output": "mul_162", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_90", + "inputs": [ + "mul_162", + "p_model_model_language_model_layers_11_mlp_gate_proj_weight", + "p_model_model_language_model_layers_11_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_90", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_91", + "inputs": [ + "mul_162", + "p_model_model_language_model_layers_11_mlp_up_proj_weight", + "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_91", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_163", + "inputs": [ + "linear_90", + "linear_91" + ], + "output": "mul_163", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_92", + "inputs": [ + "mul_163", + "p_model_model_language_model_layers_11_mlp_down_proj_weight", + "p_model_model_language_model_layers_11_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_92", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_128", + "inputs": [ + "add_125", + "linear_92" + ], + "output": "add_128", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_165", + "inputs": [ + "add_128", + "p_model_model_language_model_layers_12_input_layernorm_weight" + ], + "output": "mul_165", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_93", + "inputs": [ + "mul_165", + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_93", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_55", + "inputs": [ + "linear_93" + ], + "output": "transpose_55", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_94", + "inputs": [ + "mul_165", + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_94", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_95", + "inputs": [ + "mul_165", + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_95", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_96", + "inputs": [ + "mul_165", + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_96", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_22", + "inputs": [ + "states_24", + "transpose_55" + ], + "output": "cat_22", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_112", + "inputs": [ + "cat_22" + ], + "output": "slice_112", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_24" + }, + { + "node": "conv1d_9", + "inputs": [ + "cat_22", + "p_model_model_language_model_layers_12_linear_attn_conv1d_weight" + ], + "output": "conv1d_9", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_113", + "inputs": [ + "conv1d_9" + ], + "output": "slice_113", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_30", + "inputs": [ + "slice_113" + ], + "output": "silu_30", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_114", + "inputs": [ + "silu_30" + ], + "output": "slice_114", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_56", + "inputs": [ + "slice_114" + ], + "output": "transpose_56", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_115", + "inputs": [ + "transpose_56" + ], + "output": "slice_115", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_116", + "inputs": [ + "transpose_56" + ], + "output": "slice_116", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_117", + "inputs": [ + "transpose_56" + ], + "output": "slice_117", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_12", + "inputs": [ + "linear_95" + ], + "output": "sigmoid_12", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_9", + "inputs": [ + "p_model_model_language_model_layers_12_linear_attn_a_log" + ], + "output": "exp_9", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_15", + "inputs": [ + "exp_9" + ], + "output": "neg_15", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_131", + "inputs": [ + "linear_96", + "p_model_model_language_model_layers_12_linear_attn_dt_bias" + ], + "output": "add_131", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_9", + "inputs": [ + "add_131" + ], + "output": "softplus_9", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_166", + "inputs": [ + "neg_15", + "softplus_9" + ], + "output": "mul_166", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_167", + "inputs": [ + "slice_115" + ], + "output": "mul_167", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_19", + "inputs": [ + "mul_167" + ], + "output": "sum_19", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_132", + "inputs": [ + "sum_19" + ], + "output": "add_132", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_58", + "inputs": [ + "add_132" + ], + "output": "rsqrt_58", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_168", + "inputs": [ + "slice_115", + "rsqrt_58" + ], + "output": "mul_168", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_169", + "inputs": [ + "slice_116" + ], + "output": "mul_169", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_20", + "inputs": [ + "mul_169" + ], + "output": "sum_20", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_133", + "inputs": [ + "sum_20" + ], + "output": "add_133", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_59", + "inputs": [ + "add_133" + ], + "output": "rsqrt_59", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_170", + "inputs": [ + "slice_116", + "rsqrt_59" + ], + "output": "mul_170", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_9", + "inputs": [ + "mul_168", + "mul_170", + "slice_117", + "mul_166", + "sigmoid_12", + "states_25" + ], + "output": "normalized_gated_delta_rule_9", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_118", + "inputs": [ + "normalized_gated_delta_rule_9" + ], + "output": "slice_118", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_25" + }, + { + "node": "slice_119", + "inputs": [ + "normalized_gated_delta_rule_9" + ], + "output": "slice_119", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_57", + "inputs": [ + "slice_119" + ], + "output": "transpose_57", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_172", + "inputs": [ + "transpose_57", + "p_model_model_language_model_layers_12_linear_attn_norm_weight" + ], + "output": "mul_172", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_173", + "inputs": [ + "linear_94", + "mul_172" + ], + "output": "mul_173", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_97", + "inputs": [ + "mul_173", + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_97", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_135", + "inputs": [ + "add_128", + "linear_97" + ], + "output": "add_135", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_175", + "inputs": [ + "add_135", + "p_model_model_language_model_layers_12_post_attention_layernorm_weight" + ], + "output": "mul_175", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_98", + "inputs": [ + "mul_175", + "p_model_model_language_model_layers_12_mlp_gate_proj_weight", + "p_model_model_language_model_layers_12_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_98", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_99", + "inputs": [ + "mul_175", + "p_model_model_language_model_layers_12_mlp_up_proj_weight", + "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_99", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_176", + "inputs": [ + "linear_98", + "linear_99" + ], + "output": "mul_176", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_100", + "inputs": [ + "mul_176", + "p_model_model_language_model_layers_12_mlp_down_proj_weight", + "p_model_model_language_model_layers_12_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_100", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_138", + "inputs": [ + "add_135", + "linear_100" + ], + "output": "add_138", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_178", + "inputs": [ + "add_138", + "p_model_model_language_model_layers_13_input_layernorm_weight" + ], + "output": "mul_178", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_101", + "inputs": [ + "mul_178", + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_101", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_58", + "inputs": [ + "linear_101" + ], + "output": "transpose_58", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_102", + "inputs": [ + "mul_178", + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_102", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_103", + "inputs": [ + "mul_178", + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_103", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_104", + "inputs": [ + "mul_178", + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_104", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_23", + "inputs": [ + "states_26", + "transpose_58" + ], + "output": "cat_23", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_120", + "inputs": [ + "cat_23" + ], + "output": "slice_120", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_26" + }, + { + "node": "conv1d_10", + "inputs": [ + "cat_23", + "p_model_model_language_model_layers_13_linear_attn_conv1d_weight" + ], + "output": "conv1d_10", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_121", + "inputs": [ + "conv1d_10" + ], + "output": "slice_121", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_33", + "inputs": [ + "slice_121" + ], + "output": "silu_33", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_122", + "inputs": [ + "silu_33" + ], + "output": "slice_122", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_59", + "inputs": [ + "slice_122" + ], + "output": "transpose_59", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_123", + "inputs": [ + "transpose_59" + ], + "output": "slice_123", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_124", + "inputs": [ + "transpose_59" + ], + "output": "slice_124", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_125", + "inputs": [ + "transpose_59" + ], + "output": "slice_125", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_13", + "inputs": [ + "linear_103" + ], + "output": "sigmoid_13", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_10", + "inputs": [ + "p_model_model_language_model_layers_13_linear_attn_a_log" + ], + "output": "exp_10", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_16", + "inputs": [ + "exp_10" + ], + "output": "neg_16", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_141", + "inputs": [ + "linear_104", + "p_model_model_language_model_layers_13_linear_attn_dt_bias" + ], + "output": "add_141", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_10", + "inputs": [ + "add_141" + ], + "output": "softplus_10", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_179", + "inputs": [ + "neg_16", + "softplus_10" + ], + "output": "mul_179", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_180", + "inputs": [ + "slice_123" + ], + "output": "mul_180", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_21", + "inputs": [ + "mul_180" + ], + "output": "sum_21", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_142", + "inputs": [ + "sum_21" + ], + "output": "add_142", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_63", + "inputs": [ + "add_142" + ], + "output": "rsqrt_63", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_181", + "inputs": [ + "slice_123", + "rsqrt_63" + ], + "output": "mul_181", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_182", + "inputs": [ + "slice_124" + ], + "output": "mul_182", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_22", + "inputs": [ + "mul_182" + ], + "output": "sum_22", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_143", + "inputs": [ + "sum_22" + ], + "output": "add_143", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_64", + "inputs": [ + "add_143" + ], + "output": "rsqrt_64", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_183", + "inputs": [ + "slice_124", + "rsqrt_64" + ], + "output": "mul_183", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_10", + "inputs": [ + "mul_181", + "mul_183", + "slice_125", + "mul_179", + "sigmoid_13", + "states_27" + ], + "output": "normalized_gated_delta_rule_10", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_126", + "inputs": [ + "normalized_gated_delta_rule_10" + ], + "output": "slice_126", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_27" + }, + { + "node": "slice_127", + "inputs": [ + "normalized_gated_delta_rule_10" + ], + "output": "slice_127", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_60", + "inputs": [ + "slice_127" + ], + "output": "transpose_60", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_185", + "inputs": [ + "transpose_60", + "p_model_model_language_model_layers_13_linear_attn_norm_weight" + ], + "output": "mul_185", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_186", + "inputs": [ + "linear_102", + "mul_185" + ], + "output": "mul_186", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_105", + "inputs": [ + "mul_186", + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_105", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_145", + "inputs": [ + "add_138", + "linear_105" + ], + "output": "add_145", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_188", + "inputs": [ + "add_145", + "p_model_model_language_model_layers_13_post_attention_layernorm_weight" + ], + "output": "mul_188", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_106", + "inputs": [ + "mul_188", + "p_model_model_language_model_layers_13_mlp_gate_proj_weight", + "p_model_model_language_model_layers_13_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_106", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_107", + "inputs": [ + "mul_188", + "p_model_model_language_model_layers_13_mlp_up_proj_weight", + "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_107", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_189", + "inputs": [ + "linear_106", + "linear_107" + ], + "output": "mul_189", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_108", + "inputs": [ + "mul_189", + "p_model_model_language_model_layers_13_mlp_down_proj_weight", + "p_model_model_language_model_layers_13_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_108", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_148", + "inputs": [ + "add_145", + "linear_108" + ], + "output": "add_148", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_191", + "inputs": [ + "add_148", + "p_model_model_language_model_layers_14_input_layernorm_weight" + ], + "output": "mul_191", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_109", + "inputs": [ + "mul_191", + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_109", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_61", + "inputs": [ + "linear_109" + ], + "output": "transpose_61", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_110", + "inputs": [ + "mul_191", + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_110", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_111", + "inputs": [ + "mul_191", + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_111", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_112", + "inputs": [ + "mul_191", + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_112", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_24", + "inputs": [ + "states_28", + "transpose_61" + ], + "output": "cat_24", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_128", + "inputs": [ + "cat_24" + ], + "output": "slice_128", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_28" + }, + { + "node": "conv1d_11", + "inputs": [ + "cat_24", + "p_model_model_language_model_layers_14_linear_attn_conv1d_weight" + ], + "output": "conv1d_11", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_129", + "inputs": [ + "conv1d_11" + ], + "output": "slice_129", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_36", + "inputs": [ + "slice_129" + ], + "output": "silu_36", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_130", + "inputs": [ + "silu_36" + ], + "output": "slice_130", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_62", + "inputs": [ + "slice_130" + ], + "output": "transpose_62", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_131", + "inputs": [ + "transpose_62" + ], + "output": "slice_131", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_132", + "inputs": [ + "transpose_62" + ], + "output": "slice_132", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_133", + "inputs": [ + "transpose_62" + ], + "output": "slice_133", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_14", + "inputs": [ + "linear_111" + ], + "output": "sigmoid_14", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_11", + "inputs": [ + "p_model_model_language_model_layers_14_linear_attn_a_log" + ], + "output": "exp_11", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_17", + "inputs": [ + "exp_11" + ], + "output": "neg_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_151", + "inputs": [ + "linear_112", + "p_model_model_language_model_layers_14_linear_attn_dt_bias" + ], + "output": "add_151", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_11", + "inputs": [ + "add_151" + ], + "output": "softplus_11", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_192", + "inputs": [ + "neg_17", + "softplus_11" + ], + "output": "mul_192", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_193", + "inputs": [ + "slice_131" + ], + "output": "mul_193", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_23", + "inputs": [ + "mul_193" + ], + "output": "sum_23", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_152", + "inputs": [ + "sum_23" + ], + "output": "add_152", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_68", + "inputs": [ + "add_152" + ], + "output": "rsqrt_68", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_194", + "inputs": [ + "slice_131", + "rsqrt_68" + ], + "output": "mul_194", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_195", + "inputs": [ + "slice_132" + ], + "output": "mul_195", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_24", + "inputs": [ + "mul_195" + ], + "output": "sum_24", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_153", + "inputs": [ + "sum_24" + ], + "output": "add_153", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_69", + "inputs": [ + "add_153" + ], + "output": "rsqrt_69", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_196", + "inputs": [ + "slice_132", + "rsqrt_69" + ], + "output": "mul_196", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_11", + "inputs": [ + "mul_194", + "mul_196", + "slice_133", + "mul_192", + "sigmoid_14", + "states_29" + ], + "output": "normalized_gated_delta_rule_11", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_134", + "inputs": [ + "normalized_gated_delta_rule_11" + ], + "output": "slice_134", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_29" + }, + { + "node": "slice_135", + "inputs": [ + "normalized_gated_delta_rule_11" + ], + "output": "slice_135", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_63", + "inputs": [ + "slice_135" + ], + "output": "transpose_63", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_198", + "inputs": [ + "transpose_63", + "p_model_model_language_model_layers_14_linear_attn_norm_weight" + ], + "output": "mul_198", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_199", + "inputs": [ + "linear_110", + "mul_198" + ], + "output": "mul_199", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_113", + "inputs": [ + "mul_199", + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_113", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_155", + "inputs": [ + "add_148", + "linear_113" + ], + "output": "add_155", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_201", + "inputs": [ + "add_155", + "p_model_model_language_model_layers_14_post_attention_layernorm_weight" + ], + "output": "mul_201", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_114", + "inputs": [ + "mul_201", + "p_model_model_language_model_layers_14_mlp_gate_proj_weight", + "p_model_model_language_model_layers_14_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_114", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_115", + "inputs": [ + "mul_201", + "p_model_model_language_model_layers_14_mlp_up_proj_weight", + "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_115", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_202", + "inputs": [ + "linear_114", + "linear_115" + ], + "output": "mul_202", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_116", + "inputs": [ + "mul_202", + "p_model_model_language_model_layers_14_mlp_down_proj_weight", + "p_model_model_language_model_layers_14_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_116", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_158", + "inputs": [ + "add_155", + "linear_116" + ], + "output": "add_158", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_204", + "inputs": [ + "add_158", + "p_model_model_language_model_layers_15_input_layernorm_weight" + ], + "output": "mul_204", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_117", + "inputs": [ + "mul_204", + "p_model_model_language_model_layers_15_self_attn_q_proj_weight", + "p_model_model_language_model_layers_15_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_117", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d1e9766c981b6c437331275b2afdfc74e0789005b52c2015839fc2394961da1b.wgsl" + }, + { + "node": "slice_136", + "inputs": [ + "linear_117" + ], + "output": "slice_136", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3a76085983a5330bf07a4166ea4975700403f9d8035a8ef450f7f18b9cf47c66.wgsl" + }, + { + "node": "slice_137", + "inputs": [ + "linear_117" + ], + "output": "slice_137", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9044193f1f4724f15ef5fb41040b7fbac5f62fcaea80594737214bf9cd6be41a.wgsl" + }, + { + "node": "mul_206", + "inputs": [ + "slice_136", + "p_model_model_language_model_layers_15_self_attn_q_norm_weight" + ], + "output": "mul_206", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/ef364f47cced448917d1b6842c708dee101d53d7a781014f9a822214cea72bf8.wgsl" + }, + { + "node": "transpose_64", + "inputs": [ + "mul_206" + ], + "output": "transpose_64", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/1d0cbb068fe17123e646934bfa1259f19e5db04e45d72a157d3ea4fa48c81aa0.wgsl" + }, + { + "node": "linear_118", + "inputs": [ + "mul_204", + "p_model_model_language_model_layers_15_self_attn_k_proj_weight", + "p_model_model_language_model_layers_15_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_118", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/c5b7564d87c89606da1cba30e05e401e4fb8c83957e779a694d2ec2c472d88ad.wgsl" + }, + { + "node": "mul_208", + "inputs": [ + "linear_118", + "p_model_model_language_model_layers_15_self_attn_k_norm_weight" + ], + "output": "mul_208", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl" + }, + { + "node": "transpose_65", + "inputs": [ + "mul_208" + ], + "output": "transpose_65", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/bcb78b92dbbfeb212c190e064bc2f9d8e79c201398cf38c9aa62df6fdba94469.wgsl" + }, + { + "node": "linear_119", + "inputs": [ + "mul_204", + "p_model_model_language_model_layers_15_self_attn_v_proj_weight", + "p_model_model_language_model_layers_15_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_119", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/c5b7564d87c89606da1cba30e05e401e4fb8c83957e779a694d2ec2c472d88ad.wgsl" + }, + { + "node": "transpose_66", + "inputs": [ + "linear_119" + ], + "output": "transpose_66", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/bcb78b92dbbfeb212c190e064bc2f9d8e79c201398cf38c9aa62df6fdba94469.wgsl" + }, + { + "node": "slice_138", + "inputs": [ + "transpose_64" + ], + "output": "slice_138", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/4d8ebffbbf193b18c13a5cc17d286bf5c92c5ba53ca022af78a9151ec1e5b9c0.wgsl" + }, + { + "node": "slice_139", + "inputs": [ + "transpose_64" + ], + "output": "slice_139", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/4406ab1e6e4b9eeb91ace297320aa05a55cc481e02894a4a74b655b4070ade36.wgsl" + }, + { + "node": "slice_140", + "inputs": [ + "transpose_65" + ], + "output": "slice_140", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/89f98b79c18e7f4bbbd79cbd8d3e80108d0e1e4407c161d9637b6a88d3fbe06b.wgsl" + }, + { + "node": "slice_141", + "inputs": [ + "transpose_65" + ], + "output": "slice_141", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd70e4fc42ac55f0f52612841bcc3f9fe7df5bbb0fc76f94e56a131a92454658.wgsl" + }, + { + "node": "mul_209", + "inputs": [ + "slice_138", + "mul" + ], + "output": "mul_209", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl" + }, + { + "node": "slice_142", + "inputs": [ + "slice_138" + ], + "output": "slice_142", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/2c66dcac8fdd3caa9ff1e2441524070c21c24df9526bee44c49e765a8d4c8222.wgsl" + }, + { + "node": "slice_143", + "inputs": [ + "slice_138" + ], + "output": "slice_143", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/33e8ec04c66adb4988f562f0f9d3e92d29dfda91a56d0ed62be598c7c737f5ba.wgsl" + }, + { + "node": "neg_18", + "inputs": [ + "slice_143" + ], + "output": "neg_18", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/3e34409407100d4858c4c109cf62f3eaf2f4c6081166b0ec8679105e6dfbe7fe.wgsl" + }, + { + "node": "cat_25", + "inputs": [ + "neg_18", + "slice_142" + ], + "output": "cat_25", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/85964f8bbe5a2e4d430930e1050ea76871f0f2453c3cd4ba020002c0562ca680.wgsl" + }, + { + "node": "mul_210", + "inputs": [ + "cat_25", + "mul_1" + ], + "output": "mul_210", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl" + }, + { + "node": "add_165", + "inputs": [ + "mul_209", + "mul_210" + ], + "output": "add_165", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/d2866fd21f58839ea7d43bf83fde4243c38e4f1771034520abb020be5d819a96.wgsl" + }, + { + "node": "mul_211", + "inputs": [ + "slice_140", + "mul" + ], + "output": "mul_211", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl" + }, + { + "node": "slice_144", + "inputs": [ + "slice_140" + ], + "output": "slice_144", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/7b7c42c04ea3c6695ddead641e88b82d8f130015e138fc323eec509b5b56cfa9.wgsl" + }, + { + "node": "slice_145", + "inputs": [ + "slice_140" + ], + "output": "slice_145", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/6d82b7ed99a04f03fa0b9d241188d9613fa3ce9ff425cf8ee30a7f4ce6ded13d.wgsl" + }, + { + "node": "neg_19", + "inputs": [ + "slice_145" + ], + "output": "neg_19", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/adf1e58a124e1301625024ae142dca5fb1dd50799fa94b52eaeffb82ea99852c.wgsl" + }, + { + "node": "cat_26", + "inputs": [ + "neg_19", + "slice_144" + ], + "output": "cat_26", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl" + }, + { + "node": "mul_212", + "inputs": [ + "cat_26", + "mul_1" + ], + "output": "mul_212", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl" + }, + { + "node": "add_166", + "inputs": [ + "mul_211", + "mul_212" + ], + "output": "add_166", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl" + }, + { + "node": "cat_27", + "inputs": [ + "add_165", + "slice_139" + ], + "output": "cat_27", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/8affc7c0e85127430e406be35ebb5c5354a8dd26ecc4ae9ab6ace80ce6775f46.wgsl" + }, + { + "node": "cat_28", + "inputs": [ + "add_166", + "slice_141" + ], + "output": "cat_28", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl" + }, + { + "node": "cache_update_6", + "inputs": [ + "cat_28", + "position_ids" + ], + "output": "cache_update_6", + "workgroups": [ + 512, + 1, + 1 + ], + "inPlaceOf": "states_30", + "kernel": "kernels/e369b5c120d84bfebbc35e3da096c33e55d20a607aecd5d82da5de4c95176e00.wgsl" + }, + { + "node": "cache_update_7", + "inputs": [ + "transpose_66", + "position_ids" + ], + "output": "cache_update_7", + "workgroups": [ + 512, + 1, + 1 + ], + "inPlaceOf": "states_31", + "kernel": "kernels/e369b5c120d84bfebbc35e3da096c33e55d20a607aecd5d82da5de4c95176e00.wgsl" + }, + { + "node": "matmul_8", + "inputs": [ + "cat_27", + "cache_update_6", + "cache_update_7", + "attention_mask" + ], + "output": "matmul_8", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/140ebe12d31f1420a8a159b91ae1773390190f496e06e0e5d7b5351f38245068.wgsl" + }, + { + "node": "transpose_68", + "inputs": [ + "matmul_8" + ], + "output": "transpose_68", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4f2c0d08150d4b97d90c244db2d41baca81484c2630dbbd6d586f2d8a4b07b8b.wgsl" + }, + { + "node": "sigmoid_15", + "inputs": [ + "slice_137" + ], + "output": "sigmoid_15", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/50535e4c99d9fb4a272d8260a1415f6b38a7ad1025da64e145f230df697d9b73.wgsl" + }, + { + "node": "mul_214", + "inputs": [ + "transpose_68", + "sigmoid_15" + ], + "output": "mul_214", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/a9f6bc361f1ea31f79f0ef6164872eeb04896a45338871a2d779789b47b1920b.wgsl" + }, + { + "node": "linear_120", + "inputs": [ + "mul_214", + "p_model_model_language_model_layers_15_self_attn_o_proj_weight", + "p_model_model_language_model_layers_15_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_120", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_168", + "inputs": [ + "add_158", + "linear_120" + ], + "output": "add_168", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_216", + "inputs": [ + "add_168", + "p_model_model_language_model_layers_15_post_attention_layernorm_weight" + ], + "output": "mul_216", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_121", + "inputs": [ + "mul_216", + "p_model_model_language_model_layers_15_mlp_gate_proj_weight", + "p_model_model_language_model_layers_15_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_121", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_122", + "inputs": [ + "mul_216", + "p_model_model_language_model_layers_15_mlp_up_proj_weight", + "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_122", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_217", + "inputs": [ + "linear_121", + "linear_122" + ], + "output": "mul_217", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_123", + "inputs": [ + "mul_217", + "p_model_model_language_model_layers_15_mlp_down_proj_weight", + "p_model_model_language_model_layers_15_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_123", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_171", + "inputs": [ + "add_168", + "linear_123" + ], + "output": "add_171", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_219", + "inputs": [ + "add_171", + "p_model_model_language_model_layers_16_input_layernorm_weight" + ], + "output": "mul_219", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_124", + "inputs": [ + "mul_219", + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_124", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_69", + "inputs": [ + "linear_124" + ], + "output": "transpose_69", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_125", + "inputs": [ + "mul_219", + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_125", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_126", + "inputs": [ + "mul_219", + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_126", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_127", + "inputs": [ + "mul_219", + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_127", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_29", + "inputs": [ + "states_32", + "transpose_69" + ], + "output": "cat_29", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_146", + "inputs": [ + "cat_29" + ], + "output": "slice_146", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_32" + }, + { + "node": "conv1d_12", + "inputs": [ + "cat_29", + "p_model_model_language_model_layers_16_linear_attn_conv1d_weight" + ], + "output": "conv1d_12", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_147", + "inputs": [ + "conv1d_12" + ], + "output": "slice_147", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_40", + "inputs": [ + "slice_147" + ], + "output": "silu_40", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_148", + "inputs": [ + "silu_40" + ], + "output": "slice_148", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_70", + "inputs": [ + "slice_148" + ], + "output": "transpose_70", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_149", + "inputs": [ + "transpose_70" + ], + "output": "slice_149", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_150", + "inputs": [ + "transpose_70" + ], + "output": "slice_150", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_151", + "inputs": [ + "transpose_70" + ], + "output": "slice_151", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_16", + "inputs": [ + "linear_126" + ], + "output": "sigmoid_16", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_12", + "inputs": [ + "p_model_model_language_model_layers_16_linear_attn_a_log" + ], + "output": "exp_12", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_20", + "inputs": [ + "exp_12" + ], + "output": "neg_20", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_174", + "inputs": [ + "linear_127", + "p_model_model_language_model_layers_16_linear_attn_dt_bias" + ], + "output": "add_174", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_12", + "inputs": [ + "add_174" + ], + "output": "softplus_12", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_220", + "inputs": [ + "neg_20", + "softplus_12" + ], + "output": "mul_220", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_221", + "inputs": [ + "slice_149" + ], + "output": "mul_221", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_25", + "inputs": [ + "mul_221" + ], + "output": "sum_25", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_175", + "inputs": [ + "sum_25" + ], + "output": "add_175", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_77", + "inputs": [ + "add_175" + ], + "output": "rsqrt_77", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_222", + "inputs": [ + "slice_149", + "rsqrt_77" + ], + "output": "mul_222", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_223", + "inputs": [ + "slice_150" + ], + "output": "mul_223", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_26", + "inputs": [ + "mul_223" + ], + "output": "sum_26", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_176", + "inputs": [ + "sum_26" + ], + "output": "add_176", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_78", + "inputs": [ + "add_176" + ], + "output": "rsqrt_78", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_224", + "inputs": [ + "slice_150", + "rsqrt_78" + ], + "output": "mul_224", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_12", + "inputs": [ + "mul_222", + "mul_224", + "slice_151", + "mul_220", + "sigmoid_16", + "states_33" + ], + "output": "normalized_gated_delta_rule_12", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_152", + "inputs": [ + "normalized_gated_delta_rule_12" + ], + "output": "slice_152", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_33" + }, + { + "node": "slice_153", + "inputs": [ + "normalized_gated_delta_rule_12" + ], + "output": "slice_153", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_71", + "inputs": [ + "slice_153" + ], + "output": "transpose_71", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_226", + "inputs": [ + "transpose_71", + "p_model_model_language_model_layers_16_linear_attn_norm_weight" + ], + "output": "mul_226", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_227", + "inputs": [ + "linear_125", + "mul_226" + ], + "output": "mul_227", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_128", + "inputs": [ + "mul_227", + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_128", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_178", + "inputs": [ + "add_171", + "linear_128" + ], + "output": "add_178", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_229", + "inputs": [ + "add_178", + "p_model_model_language_model_layers_16_post_attention_layernorm_weight" + ], + "output": "mul_229", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_129", + "inputs": [ + "mul_229", + "p_model_model_language_model_layers_16_mlp_gate_proj_weight", + "p_model_model_language_model_layers_16_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_129", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_130", + "inputs": [ + "mul_229", + "p_model_model_language_model_layers_16_mlp_up_proj_weight", + "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_130", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_230", + "inputs": [ + "linear_129", + "linear_130" + ], + "output": "mul_230", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_131", + "inputs": [ + "mul_230", + "p_model_model_language_model_layers_16_mlp_down_proj_weight", + "p_model_model_language_model_layers_16_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_131", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_181", + "inputs": [ + "add_178", + "linear_131" + ], + "output": "add_181", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_232", + "inputs": [ + "add_181", + "p_model_model_language_model_layers_17_input_layernorm_weight" + ], + "output": "mul_232", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_132", + "inputs": [ + "mul_232", + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_132", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_72", + "inputs": [ + "linear_132" + ], + "output": "transpose_72", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_133", + "inputs": [ + "mul_232", + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_133", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_134", + "inputs": [ + "mul_232", + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_134", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_135", + "inputs": [ + "mul_232", + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_135", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_30", + "inputs": [ + "states_34", + "transpose_72" + ], + "output": "cat_30", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_154", + "inputs": [ + "cat_30" + ], + "output": "slice_154", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_34" + }, + { + "node": "conv1d_13", + "inputs": [ + "cat_30", + "p_model_model_language_model_layers_17_linear_attn_conv1d_weight" + ], + "output": "conv1d_13", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_155", + "inputs": [ + "conv1d_13" + ], + "output": "slice_155", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_43", + "inputs": [ + "slice_155" + ], + "output": "silu_43", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_156", + "inputs": [ + "silu_43" + ], + "output": "slice_156", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_73", + "inputs": [ + "slice_156" + ], + "output": "transpose_73", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_157", + "inputs": [ + "transpose_73" + ], + "output": "slice_157", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_158", + "inputs": [ + "transpose_73" + ], + "output": "slice_158", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_159", + "inputs": [ + "transpose_73" + ], + "output": "slice_159", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_17", + "inputs": [ + "linear_134" + ], + "output": "sigmoid_17", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_13", + "inputs": [ + "p_model_model_language_model_layers_17_linear_attn_a_log" + ], + "output": "exp_13", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_21", + "inputs": [ + "exp_13" + ], + "output": "neg_21", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_184", + "inputs": [ + "linear_135", + "p_model_model_language_model_layers_17_linear_attn_dt_bias" + ], + "output": "add_184", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_13", + "inputs": [ + "add_184" + ], + "output": "softplus_13", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_233", + "inputs": [ + "neg_21", + "softplus_13" + ], + "output": "mul_233", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_234", + "inputs": [ + "slice_157" + ], + "output": "mul_234", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_27", + "inputs": [ + "mul_234" + ], + "output": "sum_27", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_185", + "inputs": [ + "sum_27" + ], + "output": "add_185", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_82", + "inputs": [ + "add_185" + ], + "output": "rsqrt_82", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_235", + "inputs": [ + "slice_157", + "rsqrt_82" + ], + "output": "mul_235", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_236", + "inputs": [ + "slice_158" + ], + "output": "mul_236", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_28", + "inputs": [ + "mul_236" + ], + "output": "sum_28", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_186", + "inputs": [ + "sum_28" + ], + "output": "add_186", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_83", + "inputs": [ + "add_186" + ], + "output": "rsqrt_83", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_237", + "inputs": [ + "slice_158", + "rsqrt_83" + ], + "output": "mul_237", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_13", + "inputs": [ + "mul_235", + "mul_237", + "slice_159", + "mul_233", + "sigmoid_17", + "states_35" + ], + "output": "normalized_gated_delta_rule_13", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_160", + "inputs": [ + "normalized_gated_delta_rule_13" + ], + "output": "slice_160", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_35" + }, + { + "node": "slice_161", + "inputs": [ + "normalized_gated_delta_rule_13" + ], + "output": "slice_161", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_74", + "inputs": [ + "slice_161" + ], + "output": "transpose_74", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_239", + "inputs": [ + "transpose_74", + "p_model_model_language_model_layers_17_linear_attn_norm_weight" + ], + "output": "mul_239", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_240", + "inputs": [ + "linear_133", + "mul_239" + ], + "output": "mul_240", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_136", + "inputs": [ + "mul_240", + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_136", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_188", + "inputs": [ + "add_181", + "linear_136" + ], + "output": "add_188", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_242", + "inputs": [ + "add_188", + "p_model_model_language_model_layers_17_post_attention_layernorm_weight" + ], + "output": "mul_242", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_137", + "inputs": [ + "mul_242", + "p_model_model_language_model_layers_17_mlp_gate_proj_weight", + "p_model_model_language_model_layers_17_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_137", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_138", + "inputs": [ + "mul_242", + "p_model_model_language_model_layers_17_mlp_up_proj_weight", + "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_138", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_243", + "inputs": [ + "linear_137", + "linear_138" + ], + "output": "mul_243", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_139", + "inputs": [ + "mul_243", + "p_model_model_language_model_layers_17_mlp_down_proj_weight", + "p_model_model_language_model_layers_17_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_139", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_191", + "inputs": [ + "add_188", + "linear_139" + ], + "output": "add_191", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_245", + "inputs": [ + "add_191", + "p_model_model_language_model_layers_18_input_layernorm_weight" + ], + "output": "mul_245", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_140", + "inputs": [ + "mul_245", + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_140", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_75", + "inputs": [ + "linear_140" + ], + "output": "transpose_75", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_141", + "inputs": [ + "mul_245", + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_141", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_142", + "inputs": [ + "mul_245", + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_142", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_143", + "inputs": [ + "mul_245", + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_143", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_31", + "inputs": [ + "states_36", + "transpose_75" + ], + "output": "cat_31", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_162", + "inputs": [ + "cat_31" + ], + "output": "slice_162", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_36" + }, + { + "node": "conv1d_14", + "inputs": [ + "cat_31", + "p_model_model_language_model_layers_18_linear_attn_conv1d_weight" + ], + "output": "conv1d_14", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_163", + "inputs": [ + "conv1d_14" + ], + "output": "slice_163", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_46", + "inputs": [ + "slice_163" + ], + "output": "silu_46", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_164", + "inputs": [ + "silu_46" + ], + "output": "slice_164", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_76", + "inputs": [ + "slice_164" + ], + "output": "transpose_76", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_165", + "inputs": [ + "transpose_76" + ], + "output": "slice_165", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_166", + "inputs": [ + "transpose_76" + ], + "output": "slice_166", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_167", + "inputs": [ + "transpose_76" + ], + "output": "slice_167", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_18", + "inputs": [ + "linear_142" + ], + "output": "sigmoid_18", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_14", + "inputs": [ + "p_model_model_language_model_layers_18_linear_attn_a_log" + ], + "output": "exp_14", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_22", + "inputs": [ + "exp_14" + ], + "output": "neg_22", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_194", + "inputs": [ + "linear_143", + "p_model_model_language_model_layers_18_linear_attn_dt_bias" + ], + "output": "add_194", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_14", + "inputs": [ + "add_194" + ], + "output": "softplus_14", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_246", + "inputs": [ + "neg_22", + "softplus_14" + ], + "output": "mul_246", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_247", + "inputs": [ + "slice_165" + ], + "output": "mul_247", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_29", + "inputs": [ + "mul_247" + ], + "output": "sum_29", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_195", + "inputs": [ + "sum_29" + ], + "output": "add_195", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_87", + "inputs": [ + "add_195" + ], + "output": "rsqrt_87", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_248", + "inputs": [ + "slice_165", + "rsqrt_87" + ], + "output": "mul_248", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_249", + "inputs": [ + "slice_166" + ], + "output": "mul_249", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_30", + "inputs": [ + "mul_249" + ], + "output": "sum_30", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_196", + "inputs": [ + "sum_30" + ], + "output": "add_196", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_88", + "inputs": [ + "add_196" + ], + "output": "rsqrt_88", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_250", + "inputs": [ + "slice_166", + "rsqrt_88" + ], + "output": "mul_250", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_14", + "inputs": [ + "mul_248", + "mul_250", + "slice_167", + "mul_246", + "sigmoid_18", + "states_37" + ], + "output": "normalized_gated_delta_rule_14", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_168", + "inputs": [ + "normalized_gated_delta_rule_14" + ], + "output": "slice_168", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_37" + }, + { + "node": "slice_169", + "inputs": [ + "normalized_gated_delta_rule_14" + ], + "output": "slice_169", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_77", + "inputs": [ + "slice_169" + ], + "output": "transpose_77", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_252", + "inputs": [ + "transpose_77", + "p_model_model_language_model_layers_18_linear_attn_norm_weight" + ], + "output": "mul_252", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_253", + "inputs": [ + "linear_141", + "mul_252" + ], + "output": "mul_253", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_144", + "inputs": [ + "mul_253", + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_144", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_198", + "inputs": [ + "add_191", + "linear_144" + ], + "output": "add_198", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_255", + "inputs": [ + "add_198", + "p_model_model_language_model_layers_18_post_attention_layernorm_weight" + ], + "output": "mul_255", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_145", + "inputs": [ + "mul_255", + "p_model_model_language_model_layers_18_mlp_gate_proj_weight", + "p_model_model_language_model_layers_18_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_145", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_146", + "inputs": [ + "mul_255", + "p_model_model_language_model_layers_18_mlp_up_proj_weight", + "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_146", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_256", + "inputs": [ + "linear_145", + "linear_146" + ], + "output": "mul_256", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_147", + "inputs": [ + "mul_256", + "p_model_model_language_model_layers_18_mlp_down_proj_weight", + "p_model_model_language_model_layers_18_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_147", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_201", + "inputs": [ + "add_198", + "linear_147" + ], + "output": "add_201", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_258", + "inputs": [ + "add_201", + "p_model_model_language_model_layers_19_input_layernorm_weight" + ], + "output": "mul_258", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_148", + "inputs": [ + "mul_258", + "p_model_model_language_model_layers_19_self_attn_q_proj_weight", + "p_model_model_language_model_layers_19_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_148", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d1e9766c981b6c437331275b2afdfc74e0789005b52c2015839fc2394961da1b.wgsl" + }, + { + "node": "slice_170", + "inputs": [ + "linear_148" + ], + "output": "slice_170", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3a76085983a5330bf07a4166ea4975700403f9d8035a8ef450f7f18b9cf47c66.wgsl" + }, + { + "node": "slice_171", + "inputs": [ + "linear_148" + ], + "output": "slice_171", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9044193f1f4724f15ef5fb41040b7fbac5f62fcaea80594737214bf9cd6be41a.wgsl" + }, + { + "node": "mul_260", + "inputs": [ + "slice_170", + "p_model_model_language_model_layers_19_self_attn_q_norm_weight" + ], + "output": "mul_260", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/ef364f47cced448917d1b6842c708dee101d53d7a781014f9a822214cea72bf8.wgsl" + }, + { + "node": "transpose_78", + "inputs": [ + "mul_260" + ], + "output": "transpose_78", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/1d0cbb068fe17123e646934bfa1259f19e5db04e45d72a157d3ea4fa48c81aa0.wgsl" + }, + { + "node": "linear_149", + "inputs": [ + "mul_258", + "p_model_model_language_model_layers_19_self_attn_k_proj_weight", + "p_model_model_language_model_layers_19_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_149", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/c5b7564d87c89606da1cba30e05e401e4fb8c83957e779a694d2ec2c472d88ad.wgsl" + }, + { + "node": "mul_262", + "inputs": [ + "linear_149", + "p_model_model_language_model_layers_19_self_attn_k_norm_weight" + ], + "output": "mul_262", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl" + }, + { + "node": "transpose_79", + "inputs": [ + "mul_262" + ], + "output": "transpose_79", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/bcb78b92dbbfeb212c190e064bc2f9d8e79c201398cf38c9aa62df6fdba94469.wgsl" + }, + { + "node": "linear_150", + "inputs": [ + "mul_258", + "p_model_model_language_model_layers_19_self_attn_v_proj_weight", + "p_model_model_language_model_layers_19_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_150", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/c5b7564d87c89606da1cba30e05e401e4fb8c83957e779a694d2ec2c472d88ad.wgsl" + }, + { + "node": "transpose_80", + "inputs": [ + "linear_150" + ], + "output": "transpose_80", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/bcb78b92dbbfeb212c190e064bc2f9d8e79c201398cf38c9aa62df6fdba94469.wgsl" + }, + { + "node": "slice_172", + "inputs": [ + "transpose_78" + ], + "output": "slice_172", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/4d8ebffbbf193b18c13a5cc17d286bf5c92c5ba53ca022af78a9151ec1e5b9c0.wgsl" + }, + { + "node": "slice_173", + "inputs": [ + "transpose_78" + ], + "output": "slice_173", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/4406ab1e6e4b9eeb91ace297320aa05a55cc481e02894a4a74b655b4070ade36.wgsl" + }, + { + "node": "slice_174", + "inputs": [ + "transpose_79" + ], + "output": "slice_174", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/89f98b79c18e7f4bbbd79cbd8d3e80108d0e1e4407c161d9637b6a88d3fbe06b.wgsl" + }, + { + "node": "slice_175", + "inputs": [ + "transpose_79" + ], + "output": "slice_175", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd70e4fc42ac55f0f52612841bcc3f9fe7df5bbb0fc76f94e56a131a92454658.wgsl" + }, + { + "node": "mul_263", + "inputs": [ + "slice_172", + "mul" + ], + "output": "mul_263", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl" + }, + { + "node": "slice_176", + "inputs": [ + "slice_172" + ], + "output": "slice_176", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/2c66dcac8fdd3caa9ff1e2441524070c21c24df9526bee44c49e765a8d4c8222.wgsl" + }, + { + "node": "slice_177", + "inputs": [ + "slice_172" + ], + "output": "slice_177", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/33e8ec04c66adb4988f562f0f9d3e92d29dfda91a56d0ed62be598c7c737f5ba.wgsl" + }, + { + "node": "neg_23", + "inputs": [ + "slice_177" + ], + "output": "neg_23", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/3e34409407100d4858c4c109cf62f3eaf2f4c6081166b0ec8679105e6dfbe7fe.wgsl" + }, + { + "node": "cat_32", + "inputs": [ + "neg_23", + "slice_176" + ], + "output": "cat_32", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/85964f8bbe5a2e4d430930e1050ea76871f0f2453c3cd4ba020002c0562ca680.wgsl" + }, + { + "node": "mul_264", + "inputs": [ + "cat_32", + "mul_1" + ], + "output": "mul_264", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl" + }, + { + "node": "add_208", + "inputs": [ + "mul_263", + "mul_264" + ], + "output": "add_208", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/d2866fd21f58839ea7d43bf83fde4243c38e4f1771034520abb020be5d819a96.wgsl" + }, + { + "node": "mul_265", + "inputs": [ + "slice_174", + "mul" + ], + "output": "mul_265", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl" + }, + { + "node": "slice_178", + "inputs": [ + "slice_174" + ], + "output": "slice_178", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/7b7c42c04ea3c6695ddead641e88b82d8f130015e138fc323eec509b5b56cfa9.wgsl" + }, + { + "node": "slice_179", + "inputs": [ + "slice_174" + ], + "output": "slice_179", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/6d82b7ed99a04f03fa0b9d241188d9613fa3ce9ff425cf8ee30a7f4ce6ded13d.wgsl" + }, + { + "node": "neg_24", + "inputs": [ + "slice_179" + ], + "output": "neg_24", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/adf1e58a124e1301625024ae142dca5fb1dd50799fa94b52eaeffb82ea99852c.wgsl" + }, + { + "node": "cat_33", + "inputs": [ + "neg_24", + "slice_178" + ], + "output": "cat_33", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl" + }, + { + "node": "mul_266", + "inputs": [ + "cat_33", + "mul_1" + ], + "output": "mul_266", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl" + }, + { + "node": "add_209", + "inputs": [ + "mul_265", + "mul_266" + ], + "output": "add_209", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl" + }, + { + "node": "cat_34", + "inputs": [ + "add_208", + "slice_173" + ], + "output": "cat_34", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/8affc7c0e85127430e406be35ebb5c5354a8dd26ecc4ae9ab6ace80ce6775f46.wgsl" + }, + { + "node": "cat_35", + "inputs": [ + "add_209", + "slice_175" + ], + "output": "cat_35", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl" + }, + { + "node": "cache_update_8", + "inputs": [ + "cat_35", + "position_ids" + ], + "output": "cache_update_8", + "workgroups": [ + 512, + 1, + 1 + ], + "inPlaceOf": "states_38", + "kernel": "kernels/e369b5c120d84bfebbc35e3da096c33e55d20a607aecd5d82da5de4c95176e00.wgsl" + }, + { + "node": "cache_update_9", + "inputs": [ + "transpose_80", + "position_ids" + ], + "output": "cache_update_9", + "workgroups": [ + 512, + 1, + 1 + ], + "inPlaceOf": "states_39", + "kernel": "kernels/e369b5c120d84bfebbc35e3da096c33e55d20a607aecd5d82da5de4c95176e00.wgsl" + }, + { + "node": "matmul_10", + "inputs": [ + "cat_34", + "cache_update_8", + "cache_update_9", + "attention_mask" + ], + "output": "matmul_10", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/140ebe12d31f1420a8a159b91ae1773390190f496e06e0e5d7b5351f38245068.wgsl" + }, + { + "node": "transpose_82", + "inputs": [ + "matmul_10" + ], + "output": "transpose_82", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4f2c0d08150d4b97d90c244db2d41baca81484c2630dbbd6d586f2d8a4b07b8b.wgsl" + }, + { + "node": "sigmoid_19", + "inputs": [ + "slice_171" + ], + "output": "sigmoid_19", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/50535e4c99d9fb4a272d8260a1415f6b38a7ad1025da64e145f230df697d9b73.wgsl" + }, + { + "node": "mul_268", + "inputs": [ + "transpose_82", + "sigmoid_19" + ], + "output": "mul_268", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/a9f6bc361f1ea31f79f0ef6164872eeb04896a45338871a2d779789b47b1920b.wgsl" + }, + { + "node": "linear_151", + "inputs": [ + "mul_268", + "p_model_model_language_model_layers_19_self_attn_o_proj_weight", + "p_model_model_language_model_layers_19_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_151", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_211", + "inputs": [ + "add_201", + "linear_151" + ], + "output": "add_211", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_270", + "inputs": [ + "add_211", + "p_model_model_language_model_layers_19_post_attention_layernorm_weight" + ], + "output": "mul_270", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_152", + "inputs": [ + "mul_270", + "p_model_model_language_model_layers_19_mlp_gate_proj_weight", + "p_model_model_language_model_layers_19_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_152", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_153", + "inputs": [ + "mul_270", + "p_model_model_language_model_layers_19_mlp_up_proj_weight", + "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_153", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_271", + "inputs": [ + "linear_152", + "linear_153" + ], + "output": "mul_271", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_154", + "inputs": [ + "mul_271", + "p_model_model_language_model_layers_19_mlp_down_proj_weight", + "p_model_model_language_model_layers_19_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_154", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_214", + "inputs": [ + "add_211", + "linear_154" + ], + "output": "add_214", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_273", + "inputs": [ + "add_214", + "p_model_model_language_model_layers_20_input_layernorm_weight" + ], + "output": "mul_273", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_155", + "inputs": [ + "mul_273", + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_155", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_83", + "inputs": [ + "linear_155" + ], + "output": "transpose_83", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_156", + "inputs": [ + "mul_273", + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_156", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_157", + "inputs": [ + "mul_273", + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_157", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_158", + "inputs": [ + "mul_273", + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_158", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_36", + "inputs": [ + "states_40", + "transpose_83" + ], + "output": "cat_36", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_180", + "inputs": [ + "cat_36" + ], + "output": "slice_180", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_40" + }, + { + "node": "conv1d_15", + "inputs": [ + "cat_36", + "p_model_model_language_model_layers_20_linear_attn_conv1d_weight" + ], + "output": "conv1d_15", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_181", + "inputs": [ + "conv1d_15" + ], + "output": "slice_181", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_50", + "inputs": [ + "slice_181" + ], + "output": "silu_50", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_182", + "inputs": [ + "silu_50" + ], + "output": "slice_182", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_84", + "inputs": [ + "slice_182" + ], + "output": "transpose_84", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_183", + "inputs": [ + "transpose_84" + ], + "output": "slice_183", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_184", + "inputs": [ + "transpose_84" + ], + "output": "slice_184", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_185", + "inputs": [ + "transpose_84" + ], + "output": "slice_185", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_20", + "inputs": [ + "linear_157" + ], + "output": "sigmoid_20", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_15", + "inputs": [ + "p_model_model_language_model_layers_20_linear_attn_a_log" + ], + "output": "exp_15", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_25", + "inputs": [ + "exp_15" + ], + "output": "neg_25", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_217", + "inputs": [ + "linear_158", + "p_model_model_language_model_layers_20_linear_attn_dt_bias" + ], + "output": "add_217", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_15", + "inputs": [ + "add_217" + ], + "output": "softplus_15", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_274", + "inputs": [ + "neg_25", + "softplus_15" + ], + "output": "mul_274", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_275", + "inputs": [ + "slice_183" + ], + "output": "mul_275", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_31", + "inputs": [ + "mul_275" + ], + "output": "sum_31", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_218", + "inputs": [ + "sum_31" + ], + "output": "add_218", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_96", + "inputs": [ + "add_218" + ], + "output": "rsqrt_96", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_276", + "inputs": [ + "slice_183", + "rsqrt_96" + ], + "output": "mul_276", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_277", + "inputs": [ + "slice_184" + ], + "output": "mul_277", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_32", + "inputs": [ + "mul_277" + ], + "output": "sum_32", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_219", + "inputs": [ + "sum_32" + ], + "output": "add_219", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_97", + "inputs": [ + "add_219" + ], + "output": "rsqrt_97", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_278", + "inputs": [ + "slice_184", + "rsqrt_97" + ], + "output": "mul_278", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_15", + "inputs": [ + "mul_276", + "mul_278", + "slice_185", + "mul_274", + "sigmoid_20", + "states_41" + ], + "output": "normalized_gated_delta_rule_15", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_186", + "inputs": [ + "normalized_gated_delta_rule_15" + ], + "output": "slice_186", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_41" + }, + { + "node": "slice_187", + "inputs": [ + "normalized_gated_delta_rule_15" + ], + "output": "slice_187", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_85", + "inputs": [ + "slice_187" + ], + "output": "transpose_85", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_280", + "inputs": [ + "transpose_85", + "p_model_model_language_model_layers_20_linear_attn_norm_weight" + ], + "output": "mul_280", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_281", + "inputs": [ + "linear_156", + "mul_280" + ], + "output": "mul_281", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_159", + "inputs": [ + "mul_281", + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_159", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_221", + "inputs": [ + "add_214", + "linear_159" + ], + "output": "add_221", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_283", + "inputs": [ + "add_221", + "p_model_model_language_model_layers_20_post_attention_layernorm_weight" + ], + "output": "mul_283", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_160", + "inputs": [ + "mul_283", + "p_model_model_language_model_layers_20_mlp_gate_proj_weight", + "p_model_model_language_model_layers_20_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_160", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_161", + "inputs": [ + "mul_283", + "p_model_model_language_model_layers_20_mlp_up_proj_weight", + "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_161", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_284", + "inputs": [ + "linear_160", + "linear_161" + ], + "output": "mul_284", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_162", + "inputs": [ + "mul_284", + "p_model_model_language_model_layers_20_mlp_down_proj_weight", + "p_model_model_language_model_layers_20_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_162", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_224", + "inputs": [ + "add_221", + "linear_162" + ], + "output": "add_224", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_286", + "inputs": [ + "add_224", + "p_model_model_language_model_layers_21_input_layernorm_weight" + ], + "output": "mul_286", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_163", + "inputs": [ + "mul_286", + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_163", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_86", + "inputs": [ + "linear_163" + ], + "output": "transpose_86", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_164", + "inputs": [ + "mul_286", + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_164", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_165", + "inputs": [ + "mul_286", + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_165", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_166", + "inputs": [ + "mul_286", + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_166", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_37", + "inputs": [ + "states_42", + "transpose_86" + ], + "output": "cat_37", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_188", + "inputs": [ + "cat_37" + ], + "output": "slice_188", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_42" + }, + { + "node": "conv1d_16", + "inputs": [ + "cat_37", + "p_model_model_language_model_layers_21_linear_attn_conv1d_weight" + ], + "output": "conv1d_16", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_189", + "inputs": [ + "conv1d_16" + ], + "output": "slice_189", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_53", + "inputs": [ + "slice_189" + ], + "output": "silu_53", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_190", + "inputs": [ + "silu_53" + ], + "output": "slice_190", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_87", + "inputs": [ + "slice_190" + ], + "output": "transpose_87", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_191", + "inputs": [ + "transpose_87" + ], + "output": "slice_191", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_192", + "inputs": [ + "transpose_87" + ], + "output": "slice_192", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_193", + "inputs": [ + "transpose_87" + ], + "output": "slice_193", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_21", + "inputs": [ + "linear_165" + ], + "output": "sigmoid_21", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_16", + "inputs": [ + "p_model_model_language_model_layers_21_linear_attn_a_log" + ], + "output": "exp_16", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_26", + "inputs": [ + "exp_16" + ], + "output": "neg_26", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_227", + "inputs": [ + "linear_166", + "p_model_model_language_model_layers_21_linear_attn_dt_bias" + ], + "output": "add_227", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_16", + "inputs": [ + "add_227" + ], + "output": "softplus_16", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_287", + "inputs": [ + "neg_26", + "softplus_16" + ], + "output": "mul_287", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_288", + "inputs": [ + "slice_191" + ], + "output": "mul_288", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_33", + "inputs": [ + "mul_288" + ], + "output": "sum_33", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_228", + "inputs": [ + "sum_33" + ], + "output": "add_228", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_101", + "inputs": [ + "add_228" + ], + "output": "rsqrt_101", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_289", + "inputs": [ + "slice_191", + "rsqrt_101" + ], + "output": "mul_289", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_290", + "inputs": [ + "slice_192" + ], + "output": "mul_290", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_34", + "inputs": [ + "mul_290" + ], + "output": "sum_34", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_229", + "inputs": [ + "sum_34" + ], + "output": "add_229", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_102", + "inputs": [ + "add_229" + ], + "output": "rsqrt_102", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_291", + "inputs": [ + "slice_192", + "rsqrt_102" + ], + "output": "mul_291", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_16", + "inputs": [ + "mul_289", + "mul_291", + "slice_193", + "mul_287", + "sigmoid_21", + "states_43" + ], + "output": "normalized_gated_delta_rule_16", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_194", + "inputs": [ + "normalized_gated_delta_rule_16" + ], + "output": "slice_194", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_43" + }, + { + "node": "slice_195", + "inputs": [ + "normalized_gated_delta_rule_16" + ], + "output": "slice_195", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_88", + "inputs": [ + "slice_195" + ], + "output": "transpose_88", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_293", + "inputs": [ + "transpose_88", + "p_model_model_language_model_layers_21_linear_attn_norm_weight" + ], + "output": "mul_293", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_294", + "inputs": [ + "linear_164", + "mul_293" + ], + "output": "mul_294", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_167", + "inputs": [ + "mul_294", + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_167", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_231", + "inputs": [ + "add_224", + "linear_167" + ], + "output": "add_231", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_296", + "inputs": [ + "add_231", + "p_model_model_language_model_layers_21_post_attention_layernorm_weight" + ], + "output": "mul_296", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_168", + "inputs": [ + "mul_296", + "p_model_model_language_model_layers_21_mlp_gate_proj_weight", + "p_model_model_language_model_layers_21_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_168", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_169", + "inputs": [ + "mul_296", + "p_model_model_language_model_layers_21_mlp_up_proj_weight", + "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_169", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_297", + "inputs": [ + "linear_168", + "linear_169" + ], + "output": "mul_297", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_170", + "inputs": [ + "mul_297", + "p_model_model_language_model_layers_21_mlp_down_proj_weight", + "p_model_model_language_model_layers_21_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_170", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_234", + "inputs": [ + "add_231", + "linear_170" + ], + "output": "add_234", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_299", + "inputs": [ + "add_234", + "p_model_model_language_model_layers_22_input_layernorm_weight" + ], + "output": "mul_299", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_171", + "inputs": [ + "mul_299", + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight__int8_scales" + ], + "output": "linear_171", + "workgroups": [ + 96, + 1, + 1 + ], + "kernel": "kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl" + }, + { + "node": "transpose_89", + "inputs": [ + "linear_171" + ], + "output": "transpose_89", + "workgroups": [ + 384, + 4, + 1 + ], + "kernel": "kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl" + }, + { + "node": "linear_172", + "inputs": [ + "mul_299", + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight__int8_scales" + ], + "output": "linear_172", + "workgroups": [ + 32, + 1, + 1 + ], + "kernel": "kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl" + }, + { + "node": "linear_173", + "inputs": [ + "mul_299", + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight__int8_scales" + ], + "output": "linear_173", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "linear_174", + "inputs": [ + "mul_299", + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight", + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight__int8_scales" + ], + "output": "linear_174", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl" + }, + { + "node": "cat_38", + "inputs": [ + "states_44", + "transpose_89" + ], + "output": "cat_38", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl" + }, + { + "node": "slice_196", + "inputs": [ + "cat_38" + ], + "output": "slice_196", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd8a785d4fbf09918f149c7eb5f34bd25f9358bf3412e27799a7c65436c71759.wgsl", + "inPlaceOf": "states_44" + }, + { + "node": "conv1d_17", + "inputs": [ + "cat_38", + "p_model_model_language_model_layers_22_linear_attn_conv1d_weight" + ], + "output": "conv1d_17", + "workgroups": [ + 6816, + 1, + 1 + ], + "kernel": "kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl" + }, + { + "node": "slice_197", + "inputs": [ + "conv1d_17" + ], + "output": "slice_197", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl" + }, + { + "node": "silu_56", + "inputs": [ + "slice_197" + ], + "output": "silu_56", + "workgroups": [ + 6528, + 1, + 1 + ], + "kernel": "kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl" + }, + { + "node": "slice_198", + "inputs": [ + "silu_56" + ], + "output": "slice_198", + "workgroups": [ + 6144, + 1, + 1 + ], + "kernel": "kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl" + }, + { + "node": "transpose_90", + "inputs": [ + "slice_198" + ], + "output": "transpose_90", + "workgroups": [ + 4, + 384, + 1 + ], + "kernel": "kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl" + }, + { + "node": "slice_199", + "inputs": [ + "transpose_90" + ], + "output": "slice_199", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/ac173d038b5767711ad9f7cafbceb9ca3ec84aaeb3d07a3037e7cd0f83fead2c.wgsl" + }, + { + "node": "slice_200", + "inputs": [ + "transpose_90" + ], + "output": "slice_200", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl" + }, + { + "node": "slice_201", + "inputs": [ + "transpose_90" + ], + "output": "slice_201", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl" + }, + { + "node": "sigmoid_22", + "inputs": [ + "linear_173" + ], + "output": "sigmoid_22", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl" + }, + { + "node": "exp_17", + "inputs": [ + "p_model_model_language_model_layers_22_linear_attn_a_log" + ], + "output": "exp_17", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl" + }, + { + "node": "neg_27", + "inputs": [ + "exp_17" + ], + "output": "neg_27", + "workgroups": [ + 1, + 1, + 1 + ], + "kernel": "kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl" + }, + { + "node": "add_237", + "inputs": [ + "linear_174", + "p_model_model_language_model_layers_22_linear_attn_dt_bias" + ], + "output": "add_237", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl" + }, + { + "node": "softplus_17", + "inputs": [ + "add_237" + ], + "output": "softplus_17", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl" + }, + { + "node": "mul_300", + "inputs": [ + "neg_27", + "softplus_17" + ], + "output": "mul_300", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/c342e2d29647509fe0f6ac1e2dbbe74032dbe9ecd6229c03557b86f160381d2e.wgsl" + }, + { + "node": "mul_301", + "inputs": [ + "slice_199" + ], + "output": "mul_301", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_35", + "inputs": [ + "mul_301" + ], + "output": "sum_35", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_238", + "inputs": [ + "sum_35" + ], + "output": "add_238", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_106", + "inputs": [ + "add_238" + ], + "output": "rsqrt_106", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_302", + "inputs": [ + "slice_199", + "rsqrt_106" + ], + "output": "mul_302", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "mul_303", + "inputs": [ + "slice_200" + ], + "output": "mul_303", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/f5e16141122c02a1ec56f8155742fd5314ab581eac810352c06abea3f9c57690.wgsl" + }, + { + "node": "sum_36", + "inputs": [ + "mul_303" + ], + "output": "sum_36", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl" + }, + { + "node": "add_239", + "inputs": [ + "sum_36" + ], + "output": "add_239", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl" + }, + { + "node": "rsqrt_107", + "inputs": [ + "add_239" + ], + "output": "rsqrt_107", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl" + }, + { + "node": "mul_304", + "inputs": [ + "slice_200", + "rsqrt_107" + ], + "output": "mul_304", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/cd74e23e4f4177562ffb4c00641720b63d154cea3600f6177e01985d74784f37.wgsl" + }, + { + "node": "normalized_gated_delta_rule_17", + "inputs": [ + "mul_302", + "mul_304", + "slice_201", + "mul_300", + "sigmoid_22", + "states_45" + ], + "output": "normalized_gated_delta_rule_17", + "workgroups": [ + 128, + 16, + 1 + ], + "kernel": "kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl" + }, + { + "node": "slice_202", + "inputs": [ + "normalized_gated_delta_rule_17" + ], + "output": "slice_202", + "workgroups": [ + 4096, + 1, + 1 + ], + "kernel": "kernels/bb7f03e6a2277dc56084e9130086507be58871253867c0ce8ad60ee5ce4521b6.wgsl", + "inPlaceOf": "states_45" + }, + { + "node": "slice_203", + "inputs": [ + "normalized_gated_delta_rule_17" + ], + "output": "slice_203", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl" + }, + { + "node": "transpose_91", + "inputs": [ + "slice_203" + ], + "output": "transpose_91", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl" + }, + { + "node": "mul_306", + "inputs": [ + "transpose_91", + "p_model_model_language_model_layers_22_linear_attn_norm_weight" + ], + "output": "mul_306", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/c3b8a95df30cf49f5ba2a9aa459e227338c340989ea076b9cf7fc3d09ba103a4.wgsl" + }, + { + "node": "mul_307", + "inputs": [ + "linear_172", + "mul_306" + ], + "output": "mul_307", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/e1866b6254f7909c12ae9d54062b0ecf0067d708abb37c1e051ef870bd2203d7.wgsl" + }, + { + "node": "linear_175", + "inputs": [ + "mul_307", + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight", + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight__int8_scales" + ], + "output": "linear_175", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_241", + "inputs": [ + "add_234", + "linear_175" + ], + "output": "add_241", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_309", + "inputs": [ + "add_241", + "p_model_model_language_model_layers_22_post_attention_layernorm_weight" + ], + "output": "mul_309", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_176", + "inputs": [ + "mul_309", + "p_model_model_language_model_layers_22_mlp_gate_proj_weight", + "p_model_model_language_model_layers_22_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_176", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_177", + "inputs": [ + "mul_309", + "p_model_model_language_model_layers_22_mlp_up_proj_weight", + "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_177", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_310", + "inputs": [ + "linear_176", + "linear_177" + ], + "output": "mul_310", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_178", + "inputs": [ + "mul_310", + "p_model_model_language_model_layers_22_mlp_down_proj_weight", + "p_model_model_language_model_layers_22_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_178", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_244", + "inputs": [ + "add_241", + "linear_178" + ], + "output": "add_244", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_312", + "inputs": [ + "add_244", + "p_model_model_language_model_layers_23_input_layernorm_weight" + ], + "output": "mul_312", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_179", + "inputs": [ + "mul_312", + "p_model_model_language_model_layers_23_self_attn_q_proj_weight", + "p_model_model_language_model_layers_23_self_attn_q_proj_weight__int8_scales" + ], + "output": "linear_179", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d1e9766c981b6c437331275b2afdfc74e0789005b52c2015839fc2394961da1b.wgsl" + }, + { + "node": "slice_204", + "inputs": [ + "linear_179" + ], + "output": "slice_204", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/3a76085983a5330bf07a4166ea4975700403f9d8035a8ef450f7f18b9cf47c66.wgsl" + }, + { + "node": "slice_205", + "inputs": [ + "linear_179" + ], + "output": "slice_205", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9044193f1f4724f15ef5fb41040b7fbac5f62fcaea80594737214bf9cd6be41a.wgsl" + }, + { + "node": "mul_314", + "inputs": [ + "slice_204", + "p_model_model_language_model_layers_23_self_attn_q_norm_weight" + ], + "output": "mul_314", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/ef364f47cced448917d1b6842c708dee101d53d7a781014f9a822214cea72bf8.wgsl" + }, + { + "node": "transpose_92", + "inputs": [ + "mul_314" + ], + "output": "transpose_92", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/1d0cbb068fe17123e646934bfa1259f19e5db04e45d72a157d3ea4fa48c81aa0.wgsl" + }, + { + "node": "linear_180", + "inputs": [ + "mul_312", + "p_model_model_language_model_layers_23_self_attn_k_proj_weight", + "p_model_model_language_model_layers_23_self_attn_k_proj_weight__int8_scales" + ], + "output": "linear_180", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/c5b7564d87c89606da1cba30e05e401e4fb8c83957e779a694d2ec2c472d88ad.wgsl" + }, + { + "node": "mul_316", + "inputs": [ + "linear_180", + "p_model_model_language_model_layers_23_self_attn_k_norm_weight" + ], + "output": "mul_316", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl" + }, + { + "node": "transpose_93", + "inputs": [ + "mul_316" + ], + "output": "transpose_93", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/bcb78b92dbbfeb212c190e064bc2f9d8e79c201398cf38c9aa62df6fdba94469.wgsl" + }, + { + "node": "linear_181", + "inputs": [ + "mul_312", + "p_model_model_language_model_layers_23_self_attn_v_proj_weight", + "p_model_model_language_model_layers_23_self_attn_v_proj_weight__int8_scales" + ], + "output": "linear_181", + "workgroups": [ + 8, + 1, + 1 + ], + "kernel": "kernels/c5b7564d87c89606da1cba30e05e401e4fb8c83957e779a694d2ec2c472d88ad.wgsl" + }, + { + "node": "transpose_94", + "inputs": [ + "linear_181" + ], + "output": "transpose_94", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/bcb78b92dbbfeb212c190e064bc2f9d8e79c201398cf38c9aa62df6fdba94469.wgsl" + }, + { + "node": "slice_206", + "inputs": [ + "transpose_92" + ], + "output": "slice_206", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/4d8ebffbbf193b18c13a5cc17d286bf5c92c5ba53ca022af78a9151ec1e5b9c0.wgsl" + }, + { + "node": "slice_207", + "inputs": [ + "transpose_92" + ], + "output": "slice_207", + "workgroups": [ + 1536, + 1, + 1 + ], + "kernel": "kernels/4406ab1e6e4b9eeb91ace297320aa05a55cc481e02894a4a74b655b4070ade36.wgsl" + }, + { + "node": "slice_208", + "inputs": [ + "transpose_93" + ], + "output": "slice_208", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/89f98b79c18e7f4bbbd79cbd8d3e80108d0e1e4407c161d9637b6a88d3fbe06b.wgsl" + }, + { + "node": "slice_209", + "inputs": [ + "transpose_93" + ], + "output": "slice_209", + "workgroups": [ + 384, + 1, + 1 + ], + "kernel": "kernels/bd70e4fc42ac55f0f52612841bcc3f9fe7df5bbb0fc76f94e56a131a92454658.wgsl" + }, + { + "node": "mul_317", + "inputs": [ + "slice_206", + "mul" + ], + "output": "mul_317", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl" + }, + { + "node": "slice_210", + "inputs": [ + "slice_206" + ], + "output": "slice_210", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/2c66dcac8fdd3caa9ff1e2441524070c21c24df9526bee44c49e765a8d4c8222.wgsl" + }, + { + "node": "slice_211", + "inputs": [ + "slice_206" + ], + "output": "slice_211", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/33e8ec04c66adb4988f562f0f9d3e92d29dfda91a56d0ed62be598c7c737f5ba.wgsl" + }, + { + "node": "neg_28", + "inputs": [ + "slice_211" + ], + "output": "neg_28", + "workgroups": [ + 256, + 1, + 1 + ], + "kernel": "kernels/3e34409407100d4858c4c109cf62f3eaf2f4c6081166b0ec8679105e6dfbe7fe.wgsl" + }, + { + "node": "cat_39", + "inputs": [ + "neg_28", + "slice_210" + ], + "output": "cat_39", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/85964f8bbe5a2e4d430930e1050ea76871f0f2453c3cd4ba020002c0562ca680.wgsl" + }, + { + "node": "mul_318", + "inputs": [ + "cat_39", + "mul_1" + ], + "output": "mul_318", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl" + }, + { + "node": "add_251", + "inputs": [ + "mul_317", + "mul_318" + ], + "output": "add_251", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/d2866fd21f58839ea7d43bf83fde4243c38e4f1771034520abb020be5d819a96.wgsl" + }, + { + "node": "mul_319", + "inputs": [ + "slice_208", + "mul" + ], + "output": "mul_319", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl" + }, + { + "node": "slice_212", + "inputs": [ + "slice_208" + ], + "output": "slice_212", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/7b7c42c04ea3c6695ddead641e88b82d8f130015e138fc323eec509b5b56cfa9.wgsl" + }, + { + "node": "slice_213", + "inputs": [ + "slice_208" + ], + "output": "slice_213", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/6d82b7ed99a04f03fa0b9d241188d9613fa3ce9ff425cf8ee30a7f4ce6ded13d.wgsl" + }, + { + "node": "neg_29", + "inputs": [ + "slice_213" + ], + "output": "neg_29", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/adf1e58a124e1301625024ae142dca5fb1dd50799fa94b52eaeffb82ea99852c.wgsl" + }, + { + "node": "cat_40", + "inputs": [ + "neg_29", + "slice_212" + ], + "output": "cat_40", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl" + }, + { + "node": "mul_320", + "inputs": [ + "cat_40", + "mul_1" + ], + "output": "mul_320", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl" + }, + { + "node": "add_252", + "inputs": [ + "mul_319", + "mul_320" + ], + "output": "add_252", + "workgroups": [ + 128, + 1, + 1 + ], + "kernel": "kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl" + }, + { + "node": "cat_41", + "inputs": [ + "add_251", + "slice_207" + ], + "output": "cat_41", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/8affc7c0e85127430e406be35ebb5c5354a8dd26ecc4ae9ab6ace80ce6775f46.wgsl" + }, + { + "node": "cat_42", + "inputs": [ + "add_252", + "slice_209" + ], + "output": "cat_42", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl" + }, + { + "node": "cache_update_10", + "inputs": [ + "cat_42", + "position_ids" + ], + "output": "cache_update_10", + "workgroups": [ + 512, + 1, + 1 + ], + "inPlaceOf": "states_46", + "kernel": "kernels/e369b5c120d84bfebbc35e3da096c33e55d20a607aecd5d82da5de4c95176e00.wgsl" + }, + { + "node": "cache_update_11", + "inputs": [ + "transpose_94", + "position_ids" + ], + "output": "cache_update_11", + "workgroups": [ + 512, + 1, + 1 + ], + "inPlaceOf": "states_47", + "kernel": "kernels/e369b5c120d84bfebbc35e3da096c33e55d20a607aecd5d82da5de4c95176e00.wgsl" + }, + { + "node": "matmul_12", + "inputs": [ + "cat_41", + "cache_update_10", + "cache_update_11", + "attention_mask" + ], + "output": "matmul_12", + "workgroups": [ + 512, + 1, + 1 + ], + "kernel": "kernels/140ebe12d31f1420a8a159b91ae1773390190f496e06e0e5d7b5351f38245068.wgsl" + }, + { + "node": "transpose_96", + "inputs": [ + "matmul_12" + ], + "output": "transpose_96", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/4f2c0d08150d4b97d90c244db2d41baca81484c2630dbbd6d586f2d8a4b07b8b.wgsl" + }, + { + "node": "sigmoid_23", + "inputs": [ + "slice_205" + ], + "output": "sigmoid_23", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/50535e4c99d9fb4a272d8260a1415f6b38a7ad1025da64e145f230df697d9b73.wgsl" + }, + { + "node": "mul_322", + "inputs": [ + "transpose_96", + "sigmoid_23" + ], + "output": "mul_322", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/a9f6bc361f1ea31f79f0ef6164872eeb04896a45338871a2d779789b47b1920b.wgsl" + }, + { + "node": "linear_182", + "inputs": [ + "mul_322", + "p_model_model_language_model_layers_23_self_attn_o_proj_weight", + "p_model_model_language_model_layers_23_self_attn_o_proj_weight__int8_scales" + ], + "output": "linear_182", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl" + }, + { + "node": "add_254", + "inputs": [ + "add_244", + "linear_182" + ], + "output": "add_254", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_324", + "inputs": [ + "add_254", + "p_model_model_language_model_layers_23_post_attention_layernorm_weight" + ], + "output": "mul_324", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_183", + "inputs": [ + "mul_324", + "p_model_model_language_model_layers_23_mlp_gate_proj_weight", + "p_model_model_language_model_layers_23_mlp_gate_proj_weight__int4_scales" + ], + "output": "linear_183", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "linear_184", + "inputs": [ + "mul_324", + "p_model_model_language_model_layers_23_mlp_up_proj_weight", + "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales" + ], + "output": "linear_184", + "workgroups": [ + 56, + 1, + 1 + ], + "kernel": "kernels/c5a289d1fb6564db68cc05de03941dff72def270e328ba8fb520e5dd0efb075f.wgsl" + }, + { + "node": "mul_325", + "inputs": [ + "linear_183", + "linear_184" + ], + "output": "mul_325", + "workgroups": [ + 3584, + 1, + 1 + ], + "kernel": "kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl" + }, + { + "node": "linear_185", + "inputs": [ + "mul_325", + "p_model_model_language_model_layers_23_mlp_down_proj_weight", + "p_model_model_language_model_layers_23_mlp_down_proj_weight__int4_scales" + ], + "output": "linear_185", + "workgroups": [ + 16, + 1, + 1 + ], + "kernel": "kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl" + }, + { + "node": "add_257", + "inputs": [ + "add_254", + "linear_185" + ], + "output": "add_257", + "workgroups": [ + 1024, + 1, + 1 + ], + "kernel": "kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl" + }, + { + "node": "mul_327", + "inputs": [ + "add_257", + "p_model_model_language_model_norm_weight" + ], + "output": "mul_327", + "workgroups": [ + 64, + 1, + 1 + ], + "kernel": "kernels/d94fb02f4ca58bdbd2252a764fcfb703abc5d216e13eee8742d7e9f40bd0276e.wgsl" + }, + { + "node": "linear_186__part_0", + "inputs": [ + "mul_327", + "p_model_model_language_model_embed_tokens_weight__rows_0_131072", + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072" + ], + "output": "linear_186__part_0", + "workgroups": [ + 2048, + 1, + 1 + ], + "kernel": "kernels/9d9666074828f92b63a676348da9d7bb80e5a03213f1bc88f4dd8607e8a26711.wgsl" + }, + { + "node": "linear_186__part_1", + "inputs": [ + "mul_327", + "p_model_model_language_model_embed_tokens_weight__rows_131072_117248", + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248" + ], + "output": "linear_186__part_1", + "workgroups": [ + 1832, + 1, + 1 + ], + "kernel": "kernels/7559e04808def8b214f2be42df9b8309e743b79b51bb6c0b9260cb3284c92c05.wgsl" + }, + { + "node": "linear_186", + "inputs": [ + "linear_186__part_0", + "linear_186__part_1" + ], + "output": "linear_186", + "workgroups": [ + 65535, + 4, + 1 + ], + "kernel": "kernels/75984e187047e6f3dd279f8af2f7b4f6360b3afcac502b9324df6f3025cd1fd5.wgsl" + } + ], + "memory": { + "slots": [ + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 64, + 64, + 6291456, + 2097152, + 16384, + 16384, + 98304, + 512, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 4194304, + 1024, + 524288, + 1024, + 524288, + 2097152, + 4096, + 1835008, + 1835008, + 1835008, + 4096, + 128, + 256, + 256, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 98304, + 1048576, + 98304, + 1048576, + 98304, + 1048576, + 8388608, + 8388608, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 458752, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 65536, + 262144, + 524288, + 65536, + 2048, + 2048, + 786432, + 262144, + 262144, + 2048, + 2048, + 786432, + 262144, + 262144, + 134217728, + 16777216, + 120061952, + 15007744, + 262144, + 262144, + 262144, + 768, + 8192, + 8192, + 8192, + 8192, + 16384, + 1572864, + 1572864, + 1671168, + 1744896, + 524288, + 524288, + 65536, + 65536, + 131072, + 33554432, + 30015488, + 63569920 + ], + "assignment": { + "p_model_model_language_model_layers_0_input_layernorm_weight": 0, + "p_model_model_language_model_layers_0_linear_attn_a_log": 1, + "p_model_model_language_model_layers_0_linear_attn_dt_bias": 2, + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight": 3, + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight": 4, + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight": 5, + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight": 6, + "p_model_model_language_model_layers_0_linear_attn_conv1d_weight": 7, + "p_model_model_language_model_layers_0_linear_attn_norm_weight": 8, + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight": 9, + "p_model_model_language_model_layers_0_post_attention_layernorm_weight": 10, + "p_model_model_language_model_layers_0_mlp_gate_proj_weight": 11, + "p_model_model_language_model_layers_0_mlp_up_proj_weight": 12, + "p_model_model_language_model_layers_0_mlp_down_proj_weight": 13, + "p_model_model_language_model_layers_1_input_layernorm_weight": 14, + "p_model_model_language_model_layers_1_linear_attn_a_log": 15, + "p_model_model_language_model_layers_1_linear_attn_dt_bias": 16, + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight": 17, + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight": 18, + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight": 19, + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight": 20, + "p_model_model_language_model_layers_1_linear_attn_conv1d_weight": 21, + "p_model_model_language_model_layers_1_linear_attn_norm_weight": 22, + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight": 23, + "p_model_model_language_model_layers_1_post_attention_layernorm_weight": 24, + "p_model_model_language_model_layers_1_mlp_gate_proj_weight": 25, + "p_model_model_language_model_layers_1_mlp_up_proj_weight": 26, + "p_model_model_language_model_layers_1_mlp_down_proj_weight": 27, + "p_model_model_language_model_layers_2_input_layernorm_weight": 28, + "p_model_model_language_model_layers_2_linear_attn_a_log": 29, + "p_model_model_language_model_layers_2_linear_attn_dt_bias": 30, + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight": 31, + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight": 32, + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight": 33, + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight": 34, + "p_model_model_language_model_layers_2_linear_attn_conv1d_weight": 35, + "p_model_model_language_model_layers_2_linear_attn_norm_weight": 36, + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight": 37, + "p_model_model_language_model_layers_2_post_attention_layernorm_weight": 38, + "p_model_model_language_model_layers_2_mlp_gate_proj_weight": 39, + "p_model_model_language_model_layers_2_mlp_up_proj_weight": 40, + "p_model_model_language_model_layers_2_mlp_down_proj_weight": 41, + "p_model_model_language_model_layers_3_input_layernorm_weight": 42, + "p_model_model_language_model_layers_3_self_attn_q_proj_weight": 43, + "p_model_model_language_model_layers_3_self_attn_q_norm_weight": 44, + "p_model_model_language_model_layers_3_self_attn_k_proj_weight": 45, + "p_model_model_language_model_layers_3_self_attn_k_norm_weight": 46, + "p_model_model_language_model_layers_3_self_attn_v_proj_weight": 47, + "p_model_model_language_model_layers_3_self_attn_o_proj_weight": 48, + "p_model_model_language_model_layers_3_post_attention_layernorm_weight": 49, + "p_model_model_language_model_layers_3_mlp_gate_proj_weight": 50, + "p_model_model_language_model_layers_3_mlp_up_proj_weight": 51, + "p_model_model_language_model_layers_3_mlp_down_proj_weight": 52, + "p_model_model_language_model_layers_4_input_layernorm_weight": 53, + "p_model_model_language_model_layers_4_linear_attn_a_log": 54, + "p_model_model_language_model_layers_4_linear_attn_dt_bias": 55, + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight": 56, + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight": 57, + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight": 58, + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight": 59, + "p_model_model_language_model_layers_4_linear_attn_conv1d_weight": 60, + "p_model_model_language_model_layers_4_linear_attn_norm_weight": 61, + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight": 62, + "p_model_model_language_model_layers_4_post_attention_layernorm_weight": 63, + "p_model_model_language_model_layers_4_mlp_gate_proj_weight": 64, + "p_model_model_language_model_layers_4_mlp_up_proj_weight": 65, + "p_model_model_language_model_layers_4_mlp_down_proj_weight": 66, + "p_model_model_language_model_layers_5_input_layernorm_weight": 67, + "p_model_model_language_model_layers_5_linear_attn_a_log": 68, + "p_model_model_language_model_layers_5_linear_attn_dt_bias": 69, + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight": 70, + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight": 71, + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight": 72, + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight": 73, + "p_model_model_language_model_layers_5_linear_attn_conv1d_weight": 74, + "p_model_model_language_model_layers_5_linear_attn_norm_weight": 75, + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight": 76, + "p_model_model_language_model_layers_5_post_attention_layernorm_weight": 77, + "p_model_model_language_model_layers_5_mlp_gate_proj_weight": 78, + "p_model_model_language_model_layers_5_mlp_up_proj_weight": 79, + "p_model_model_language_model_layers_5_mlp_down_proj_weight": 80, + "p_model_model_language_model_layers_6_input_layernorm_weight": 81, + "p_model_model_language_model_layers_6_linear_attn_a_log": 82, + "p_model_model_language_model_layers_6_linear_attn_dt_bias": 83, + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight": 84, + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight": 85, + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight": 86, + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight": 87, + "p_model_model_language_model_layers_6_linear_attn_conv1d_weight": 88, + "p_model_model_language_model_layers_6_linear_attn_norm_weight": 89, + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight": 90, + "p_model_model_language_model_layers_6_post_attention_layernorm_weight": 91, + "p_model_model_language_model_layers_6_mlp_gate_proj_weight": 92, + "p_model_model_language_model_layers_6_mlp_up_proj_weight": 93, + "p_model_model_language_model_layers_6_mlp_down_proj_weight": 94, + "p_model_model_language_model_layers_7_input_layernorm_weight": 95, + "p_model_model_language_model_layers_7_self_attn_q_proj_weight": 96, + "p_model_model_language_model_layers_7_self_attn_q_norm_weight": 97, + "p_model_model_language_model_layers_7_self_attn_k_proj_weight": 98, + "p_model_model_language_model_layers_7_self_attn_k_norm_weight": 99, + "p_model_model_language_model_layers_7_self_attn_v_proj_weight": 100, + "p_model_model_language_model_layers_7_self_attn_o_proj_weight": 101, + "p_model_model_language_model_layers_7_post_attention_layernorm_weight": 102, + "p_model_model_language_model_layers_7_mlp_gate_proj_weight": 103, + "p_model_model_language_model_layers_7_mlp_up_proj_weight": 104, + "p_model_model_language_model_layers_7_mlp_down_proj_weight": 105, + "p_model_model_language_model_layers_8_input_layernorm_weight": 106, + "p_model_model_language_model_layers_8_linear_attn_a_log": 107, + "p_model_model_language_model_layers_8_linear_attn_dt_bias": 108, + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight": 109, + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight": 110, + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight": 111, + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight": 112, + "p_model_model_language_model_layers_8_linear_attn_conv1d_weight": 113, + "p_model_model_language_model_layers_8_linear_attn_norm_weight": 114, + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight": 115, + "p_model_model_language_model_layers_8_post_attention_layernorm_weight": 116, + "p_model_model_language_model_layers_8_mlp_gate_proj_weight": 117, + "p_model_model_language_model_layers_8_mlp_up_proj_weight": 118, + "p_model_model_language_model_layers_8_mlp_down_proj_weight": 119, + "p_model_model_language_model_layers_9_input_layernorm_weight": 120, + "p_model_model_language_model_layers_9_linear_attn_a_log": 121, + "p_model_model_language_model_layers_9_linear_attn_dt_bias": 122, + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight": 123, + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight": 124, + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight": 125, + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight": 126, + "p_model_model_language_model_layers_9_linear_attn_conv1d_weight": 127, + "p_model_model_language_model_layers_9_linear_attn_norm_weight": 128, + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight": 129, + "p_model_model_language_model_layers_9_post_attention_layernorm_weight": 130, + "p_model_model_language_model_layers_9_mlp_gate_proj_weight": 131, + "p_model_model_language_model_layers_9_mlp_up_proj_weight": 132, + "p_model_model_language_model_layers_9_mlp_down_proj_weight": 133, + "p_model_model_language_model_layers_10_input_layernorm_weight": 134, + "p_model_model_language_model_layers_10_linear_attn_a_log": 135, + "p_model_model_language_model_layers_10_linear_attn_dt_bias": 136, + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight": 137, + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight": 138, + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight": 139, + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight": 140, + "p_model_model_language_model_layers_10_linear_attn_conv1d_weight": 141, + "p_model_model_language_model_layers_10_linear_attn_norm_weight": 142, + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight": 143, + "p_model_model_language_model_layers_10_post_attention_layernorm_weight": 144, + "p_model_model_language_model_layers_10_mlp_gate_proj_weight": 145, + "p_model_model_language_model_layers_10_mlp_up_proj_weight": 146, + "p_model_model_language_model_layers_10_mlp_down_proj_weight": 147, + "p_model_model_language_model_layers_11_input_layernorm_weight": 148, + "p_model_model_language_model_layers_11_self_attn_q_proj_weight": 149, + "p_model_model_language_model_layers_11_self_attn_q_norm_weight": 150, + "p_model_model_language_model_layers_11_self_attn_k_proj_weight": 151, + "p_model_model_language_model_layers_11_self_attn_k_norm_weight": 152, + "p_model_model_language_model_layers_11_self_attn_v_proj_weight": 153, + "p_model_model_language_model_layers_11_self_attn_o_proj_weight": 154, + "p_model_model_language_model_layers_11_post_attention_layernorm_weight": 155, + "p_model_model_language_model_layers_11_mlp_gate_proj_weight": 156, + "p_model_model_language_model_layers_11_mlp_up_proj_weight": 157, + "p_model_model_language_model_layers_11_mlp_down_proj_weight": 158, + "p_model_model_language_model_layers_12_input_layernorm_weight": 159, + "p_model_model_language_model_layers_12_linear_attn_a_log": 160, + "p_model_model_language_model_layers_12_linear_attn_dt_bias": 161, + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight": 162, + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight": 163, + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight": 164, + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight": 165, + "p_model_model_language_model_layers_12_linear_attn_conv1d_weight": 166, + "p_model_model_language_model_layers_12_linear_attn_norm_weight": 167, + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight": 168, + "p_model_model_language_model_layers_12_post_attention_layernorm_weight": 169, + "p_model_model_language_model_layers_12_mlp_gate_proj_weight": 170, + "p_model_model_language_model_layers_12_mlp_up_proj_weight": 171, + "p_model_model_language_model_layers_12_mlp_down_proj_weight": 172, + "p_model_model_language_model_layers_13_input_layernorm_weight": 173, + "p_model_model_language_model_layers_13_linear_attn_a_log": 174, + "p_model_model_language_model_layers_13_linear_attn_dt_bias": 175, + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight": 176, + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight": 177, + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight": 178, + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight": 179, + "p_model_model_language_model_layers_13_linear_attn_conv1d_weight": 180, + "p_model_model_language_model_layers_13_linear_attn_norm_weight": 181, + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight": 182, + "p_model_model_language_model_layers_13_post_attention_layernorm_weight": 183, + "p_model_model_language_model_layers_13_mlp_gate_proj_weight": 184, + "p_model_model_language_model_layers_13_mlp_up_proj_weight": 185, + "p_model_model_language_model_layers_13_mlp_down_proj_weight": 186, + "p_model_model_language_model_layers_14_input_layernorm_weight": 187, + "p_model_model_language_model_layers_14_linear_attn_a_log": 188, + "p_model_model_language_model_layers_14_linear_attn_dt_bias": 189, + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight": 190, + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight": 191, + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight": 192, + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight": 193, + "p_model_model_language_model_layers_14_linear_attn_conv1d_weight": 194, + "p_model_model_language_model_layers_14_linear_attn_norm_weight": 195, + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight": 196, + "p_model_model_language_model_layers_14_post_attention_layernorm_weight": 197, + "p_model_model_language_model_layers_14_mlp_gate_proj_weight": 198, + "p_model_model_language_model_layers_14_mlp_up_proj_weight": 199, + "p_model_model_language_model_layers_14_mlp_down_proj_weight": 200, + "p_model_model_language_model_layers_15_input_layernorm_weight": 201, + "p_model_model_language_model_layers_15_self_attn_q_proj_weight": 202, + "p_model_model_language_model_layers_15_self_attn_q_norm_weight": 203, + "p_model_model_language_model_layers_15_self_attn_k_proj_weight": 204, + "p_model_model_language_model_layers_15_self_attn_k_norm_weight": 205, + "p_model_model_language_model_layers_15_self_attn_v_proj_weight": 206, + "p_model_model_language_model_layers_15_self_attn_o_proj_weight": 207, + "p_model_model_language_model_layers_15_post_attention_layernorm_weight": 208, + "p_model_model_language_model_layers_15_mlp_gate_proj_weight": 209, + "p_model_model_language_model_layers_15_mlp_up_proj_weight": 210, + "p_model_model_language_model_layers_15_mlp_down_proj_weight": 211, + "p_model_model_language_model_layers_16_input_layernorm_weight": 212, + "p_model_model_language_model_layers_16_linear_attn_a_log": 213, + "p_model_model_language_model_layers_16_linear_attn_dt_bias": 214, + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight": 215, + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight": 216, + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight": 217, + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight": 218, + "p_model_model_language_model_layers_16_linear_attn_conv1d_weight": 219, + "p_model_model_language_model_layers_16_linear_attn_norm_weight": 220, + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight": 221, + "p_model_model_language_model_layers_16_post_attention_layernorm_weight": 222, + "p_model_model_language_model_layers_16_mlp_gate_proj_weight": 223, + "p_model_model_language_model_layers_16_mlp_up_proj_weight": 224, + "p_model_model_language_model_layers_16_mlp_down_proj_weight": 225, + "p_model_model_language_model_layers_17_input_layernorm_weight": 226, + "p_model_model_language_model_layers_17_linear_attn_a_log": 227, + "p_model_model_language_model_layers_17_linear_attn_dt_bias": 228, + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight": 229, + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight": 230, + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight": 231, + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight": 232, + "p_model_model_language_model_layers_17_linear_attn_conv1d_weight": 233, + "p_model_model_language_model_layers_17_linear_attn_norm_weight": 234, + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight": 235, + "p_model_model_language_model_layers_17_post_attention_layernorm_weight": 236, + "p_model_model_language_model_layers_17_mlp_gate_proj_weight": 237, + "p_model_model_language_model_layers_17_mlp_up_proj_weight": 238, + "p_model_model_language_model_layers_17_mlp_down_proj_weight": 239, + "p_model_model_language_model_layers_18_input_layernorm_weight": 240, + "p_model_model_language_model_layers_18_linear_attn_a_log": 241, + "p_model_model_language_model_layers_18_linear_attn_dt_bias": 242, + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight": 243, + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight": 244, + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight": 245, + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight": 246, + "p_model_model_language_model_layers_18_linear_attn_conv1d_weight": 247, + "p_model_model_language_model_layers_18_linear_attn_norm_weight": 248, + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight": 249, + "p_model_model_language_model_layers_18_post_attention_layernorm_weight": 250, + "p_model_model_language_model_layers_18_mlp_gate_proj_weight": 251, + "p_model_model_language_model_layers_18_mlp_up_proj_weight": 252, + "p_model_model_language_model_layers_18_mlp_down_proj_weight": 253, + "p_model_model_language_model_layers_19_input_layernorm_weight": 254, + "p_model_model_language_model_layers_19_self_attn_q_proj_weight": 255, + "p_model_model_language_model_layers_19_self_attn_q_norm_weight": 256, + "p_model_model_language_model_layers_19_self_attn_k_proj_weight": 257, + "p_model_model_language_model_layers_19_self_attn_k_norm_weight": 258, + "p_model_model_language_model_layers_19_self_attn_v_proj_weight": 259, + "p_model_model_language_model_layers_19_self_attn_o_proj_weight": 260, + "p_model_model_language_model_layers_19_post_attention_layernorm_weight": 261, + "p_model_model_language_model_layers_19_mlp_gate_proj_weight": 262, + "p_model_model_language_model_layers_19_mlp_up_proj_weight": 263, + "p_model_model_language_model_layers_19_mlp_down_proj_weight": 264, + "p_model_model_language_model_layers_20_input_layernorm_weight": 265, + "p_model_model_language_model_layers_20_linear_attn_a_log": 266, + "p_model_model_language_model_layers_20_linear_attn_dt_bias": 267, + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight": 268, + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight": 269, + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight": 270, + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight": 271, + "p_model_model_language_model_layers_20_linear_attn_conv1d_weight": 272, + "p_model_model_language_model_layers_20_linear_attn_norm_weight": 273, + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight": 274, + "p_model_model_language_model_layers_20_post_attention_layernorm_weight": 275, + "p_model_model_language_model_layers_20_mlp_gate_proj_weight": 276, + "p_model_model_language_model_layers_20_mlp_up_proj_weight": 277, + "p_model_model_language_model_layers_20_mlp_down_proj_weight": 278, + "p_model_model_language_model_layers_21_input_layernorm_weight": 279, + "p_model_model_language_model_layers_21_linear_attn_a_log": 280, + "p_model_model_language_model_layers_21_linear_attn_dt_bias": 281, + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight": 282, + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight": 283, + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight": 284, + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight": 285, + "p_model_model_language_model_layers_21_linear_attn_conv1d_weight": 286, + "p_model_model_language_model_layers_21_linear_attn_norm_weight": 287, + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight": 288, + "p_model_model_language_model_layers_21_post_attention_layernorm_weight": 289, + "p_model_model_language_model_layers_21_mlp_gate_proj_weight": 290, + "p_model_model_language_model_layers_21_mlp_up_proj_weight": 291, + "p_model_model_language_model_layers_21_mlp_down_proj_weight": 292, + "p_model_model_language_model_layers_22_input_layernorm_weight": 293, + "p_model_model_language_model_layers_22_linear_attn_a_log": 294, + "p_model_model_language_model_layers_22_linear_attn_dt_bias": 295, + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight": 296, + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight": 297, + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight": 298, + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight": 299, + "p_model_model_language_model_layers_22_linear_attn_conv1d_weight": 300, + "p_model_model_language_model_layers_22_linear_attn_norm_weight": 301, + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight": 302, + "p_model_model_language_model_layers_22_post_attention_layernorm_weight": 303, + "p_model_model_language_model_layers_22_mlp_gate_proj_weight": 304, + "p_model_model_language_model_layers_22_mlp_up_proj_weight": 305, + "p_model_model_language_model_layers_22_mlp_down_proj_weight": 306, + "p_model_model_language_model_layers_23_input_layernorm_weight": 307, + "p_model_model_language_model_layers_23_self_attn_q_proj_weight": 308, + "p_model_model_language_model_layers_23_self_attn_q_norm_weight": 309, + "p_model_model_language_model_layers_23_self_attn_k_proj_weight": 310, + "p_model_model_language_model_layers_23_self_attn_k_norm_weight": 311, + "p_model_model_language_model_layers_23_self_attn_v_proj_weight": 312, + "p_model_model_language_model_layers_23_self_attn_o_proj_weight": 313, + "p_model_model_language_model_layers_23_post_attention_layernorm_weight": 314, + "p_model_model_language_model_layers_23_mlp_gate_proj_weight": 315, + "p_model_model_language_model_layers_23_mlp_up_proj_weight": 316, + "p_model_model_language_model_layers_23_mlp_down_proj_weight": 317, + "p_model_model_language_model_norm_weight": 318, + "b_model_model_language_model_rotary_emb_inv_freq": 319, + "input_ids": 320, + "position_ids": 321, + "attention_mask": 322, + "states_0": 323, + "states_1": 324, + "states_2": 325, + "states_3": 326, + "states_4": 327, + "states_5": 328, + "states_6": 329, + "states_7": 330, + "states_8": 331, + "states_9": 332, + "states_10": 333, + "states_11": 334, + "states_12": 335, + "states_13": 336, + "states_14": 337, + "states_15": 338, + "states_16": 339, + "states_17": 340, + "states_18": 341, + "states_19": 342, + "states_20": 343, + "states_21": 344, + "states_22": 345, + "states_23": 346, + "states_24": 347, + "states_25": 348, + "states_26": 349, + "states_27": 350, + "states_28": 351, + "states_29": 352, + "states_30": 353, + "states_31": 354, + "states_32": 355, + "states_33": 356, + "states_34": 357, + "states_35": 358, + "states_36": 359, + "states_37": 360, + "states_38": 361, + "states_39": 362, + "states_40": 363, + "states_41": 364, + "states_42": 365, + "states_43": 366, + "states_44": 367, + "states_45": 368, + "states_46": 369, + "states_47": 370, + "p_model_model_language_model_layers_0_mlp_down_proj_weight__int4_scales": 371, + "p_model_model_language_model_layers_0_mlp_gate_proj_weight__int4_scales": 372, + "p_model_model_language_model_layers_0_mlp_up_proj_weight__int4_scales": 373, + "p_model_model_language_model_layers_10_mlp_down_proj_weight__int4_scales": 374, + "p_model_model_language_model_layers_10_mlp_gate_proj_weight__int4_scales": 375, + "p_model_model_language_model_layers_10_mlp_up_proj_weight__int4_scales": 376, + "p_model_model_language_model_layers_11_mlp_down_proj_weight__int4_scales": 377, + "p_model_model_language_model_layers_11_mlp_gate_proj_weight__int4_scales": 378, + "p_model_model_language_model_layers_11_mlp_up_proj_weight__int4_scales": 379, + "p_model_model_language_model_layers_12_mlp_down_proj_weight__int4_scales": 380, + "p_model_model_language_model_layers_12_mlp_gate_proj_weight__int4_scales": 381, + "p_model_model_language_model_layers_12_mlp_up_proj_weight__int4_scales": 382, + "p_model_model_language_model_layers_13_mlp_down_proj_weight__int4_scales": 383, + "p_model_model_language_model_layers_13_mlp_gate_proj_weight__int4_scales": 384, + "p_model_model_language_model_layers_13_mlp_up_proj_weight__int4_scales": 385, + "p_model_model_language_model_layers_14_mlp_down_proj_weight__int4_scales": 386, + "p_model_model_language_model_layers_14_mlp_gate_proj_weight__int4_scales": 387, + "p_model_model_language_model_layers_14_mlp_up_proj_weight__int4_scales": 388, + "p_model_model_language_model_layers_15_mlp_down_proj_weight__int4_scales": 389, + "p_model_model_language_model_layers_15_mlp_gate_proj_weight__int4_scales": 390, + "p_model_model_language_model_layers_15_mlp_up_proj_weight__int4_scales": 391, + "p_model_model_language_model_layers_16_mlp_down_proj_weight__int4_scales": 392, + "p_model_model_language_model_layers_16_mlp_gate_proj_weight__int4_scales": 393, + "p_model_model_language_model_layers_16_mlp_up_proj_weight__int4_scales": 394, + "p_model_model_language_model_layers_17_mlp_down_proj_weight__int4_scales": 395, + "p_model_model_language_model_layers_17_mlp_gate_proj_weight__int4_scales": 396, + "p_model_model_language_model_layers_17_mlp_up_proj_weight__int4_scales": 397, + "p_model_model_language_model_layers_18_mlp_down_proj_weight__int4_scales": 398, + "p_model_model_language_model_layers_18_mlp_gate_proj_weight__int4_scales": 399, + "p_model_model_language_model_layers_18_mlp_up_proj_weight__int4_scales": 400, + "p_model_model_language_model_layers_19_mlp_down_proj_weight__int4_scales": 401, + "p_model_model_language_model_layers_19_mlp_gate_proj_weight__int4_scales": 402, + "p_model_model_language_model_layers_19_mlp_up_proj_weight__int4_scales": 403, + "p_model_model_language_model_layers_1_mlp_down_proj_weight__int4_scales": 404, + "p_model_model_language_model_layers_1_mlp_gate_proj_weight__int4_scales": 405, + "p_model_model_language_model_layers_1_mlp_up_proj_weight__int4_scales": 406, + "p_model_model_language_model_layers_20_mlp_down_proj_weight__int4_scales": 407, + "p_model_model_language_model_layers_20_mlp_gate_proj_weight__int4_scales": 408, + "p_model_model_language_model_layers_20_mlp_up_proj_weight__int4_scales": 409, + "p_model_model_language_model_layers_21_mlp_down_proj_weight__int4_scales": 410, + "p_model_model_language_model_layers_21_mlp_gate_proj_weight__int4_scales": 411, + "p_model_model_language_model_layers_21_mlp_up_proj_weight__int4_scales": 412, + "p_model_model_language_model_layers_22_mlp_down_proj_weight__int4_scales": 413, + "p_model_model_language_model_layers_22_mlp_gate_proj_weight__int4_scales": 414, + "p_model_model_language_model_layers_22_mlp_up_proj_weight__int4_scales": 415, + "p_model_model_language_model_layers_23_mlp_down_proj_weight__int4_scales": 416, + "p_model_model_language_model_layers_23_mlp_gate_proj_weight__int4_scales": 417, + "p_model_model_language_model_layers_23_mlp_up_proj_weight__int4_scales": 418, + "p_model_model_language_model_layers_2_mlp_down_proj_weight__int4_scales": 419, + "p_model_model_language_model_layers_2_mlp_gate_proj_weight__int4_scales": 420, + "p_model_model_language_model_layers_2_mlp_up_proj_weight__int4_scales": 421, + "p_model_model_language_model_layers_3_mlp_down_proj_weight__int4_scales": 422, + "p_model_model_language_model_layers_3_mlp_gate_proj_weight__int4_scales": 423, + "p_model_model_language_model_layers_3_mlp_up_proj_weight__int4_scales": 424, + "p_model_model_language_model_layers_4_mlp_down_proj_weight__int4_scales": 425, + "p_model_model_language_model_layers_4_mlp_gate_proj_weight__int4_scales": 426, + "p_model_model_language_model_layers_4_mlp_up_proj_weight__int4_scales": 427, + "p_model_model_language_model_layers_5_mlp_down_proj_weight__int4_scales": 428, + "p_model_model_language_model_layers_5_mlp_gate_proj_weight__int4_scales": 429, + "p_model_model_language_model_layers_5_mlp_up_proj_weight__int4_scales": 430, + "p_model_model_language_model_layers_6_mlp_down_proj_weight__int4_scales": 431, + "p_model_model_language_model_layers_6_mlp_gate_proj_weight__int4_scales": 432, + "p_model_model_language_model_layers_6_mlp_up_proj_weight__int4_scales": 433, + "p_model_model_language_model_layers_7_mlp_down_proj_weight__int4_scales": 434, + "p_model_model_language_model_layers_7_mlp_gate_proj_weight__int4_scales": 435, + "p_model_model_language_model_layers_7_mlp_up_proj_weight__int4_scales": 436, + "p_model_model_language_model_layers_8_mlp_down_proj_weight__int4_scales": 437, + "p_model_model_language_model_layers_8_mlp_gate_proj_weight__int4_scales": 438, + "p_model_model_language_model_layers_8_mlp_up_proj_weight__int4_scales": 439, + "p_model_model_language_model_layers_9_mlp_down_proj_weight__int4_scales": 440, + "p_model_model_language_model_layers_9_mlp_gate_proj_weight__int4_scales": 441, + "p_model_model_language_model_layers_9_mlp_up_proj_weight__int4_scales": 442, + "p_model_model_language_model_layers_0_linear_attn_in_proj_a_weight__int8_scales": 443, + "p_model_model_language_model_layers_0_linear_attn_in_proj_b_weight__int8_scales": 444, + "p_model_model_language_model_layers_0_linear_attn_in_proj_qkv_weight__int8_scales": 445, + "p_model_model_language_model_layers_0_linear_attn_in_proj_z_weight__int8_scales": 446, + "p_model_model_language_model_layers_0_linear_attn_out_proj_weight__int8_scales": 447, + "p_model_model_language_model_layers_10_linear_attn_in_proj_a_weight__int8_scales": 448, + "p_model_model_language_model_layers_10_linear_attn_in_proj_b_weight__int8_scales": 449, + "p_model_model_language_model_layers_10_linear_attn_in_proj_qkv_weight__int8_scales": 450, + "p_model_model_language_model_layers_10_linear_attn_in_proj_z_weight__int8_scales": 451, + "p_model_model_language_model_layers_10_linear_attn_out_proj_weight__int8_scales": 452, + "p_model_model_language_model_layers_11_self_attn_k_proj_weight__int8_scales": 453, + "p_model_model_language_model_layers_11_self_attn_o_proj_weight__int8_scales": 454, + "p_model_model_language_model_layers_11_self_attn_q_proj_weight__int8_scales": 455, + "p_model_model_language_model_layers_11_self_attn_v_proj_weight__int8_scales": 456, + "p_model_model_language_model_layers_12_linear_attn_in_proj_a_weight__int8_scales": 457, + "p_model_model_language_model_layers_12_linear_attn_in_proj_b_weight__int8_scales": 458, + "p_model_model_language_model_layers_12_linear_attn_in_proj_qkv_weight__int8_scales": 459, + "p_model_model_language_model_layers_12_linear_attn_in_proj_z_weight__int8_scales": 460, + "p_model_model_language_model_layers_12_linear_attn_out_proj_weight__int8_scales": 461, + "p_model_model_language_model_layers_13_linear_attn_in_proj_a_weight__int8_scales": 462, + "p_model_model_language_model_layers_13_linear_attn_in_proj_b_weight__int8_scales": 463, + "p_model_model_language_model_layers_13_linear_attn_in_proj_qkv_weight__int8_scales": 464, + "p_model_model_language_model_layers_13_linear_attn_in_proj_z_weight__int8_scales": 465, + "p_model_model_language_model_layers_13_linear_attn_out_proj_weight__int8_scales": 466, + "p_model_model_language_model_layers_14_linear_attn_in_proj_a_weight__int8_scales": 467, + "p_model_model_language_model_layers_14_linear_attn_in_proj_b_weight__int8_scales": 468, + "p_model_model_language_model_layers_14_linear_attn_in_proj_qkv_weight__int8_scales": 469, + "p_model_model_language_model_layers_14_linear_attn_in_proj_z_weight__int8_scales": 470, + "p_model_model_language_model_layers_14_linear_attn_out_proj_weight__int8_scales": 471, + "p_model_model_language_model_layers_15_self_attn_k_proj_weight__int8_scales": 472, + "p_model_model_language_model_layers_15_self_attn_o_proj_weight__int8_scales": 473, + "p_model_model_language_model_layers_15_self_attn_q_proj_weight__int8_scales": 474, + "p_model_model_language_model_layers_15_self_attn_v_proj_weight__int8_scales": 475, + "p_model_model_language_model_layers_16_linear_attn_in_proj_a_weight__int8_scales": 476, + "p_model_model_language_model_layers_16_linear_attn_in_proj_b_weight__int8_scales": 477, + "p_model_model_language_model_layers_16_linear_attn_in_proj_qkv_weight__int8_scales": 478, + "p_model_model_language_model_layers_16_linear_attn_in_proj_z_weight__int8_scales": 479, + "p_model_model_language_model_layers_16_linear_attn_out_proj_weight__int8_scales": 480, + "p_model_model_language_model_layers_17_linear_attn_in_proj_a_weight__int8_scales": 481, + "p_model_model_language_model_layers_17_linear_attn_in_proj_b_weight__int8_scales": 482, + "p_model_model_language_model_layers_17_linear_attn_in_proj_qkv_weight__int8_scales": 483, + "p_model_model_language_model_layers_17_linear_attn_in_proj_z_weight__int8_scales": 484, + "p_model_model_language_model_layers_17_linear_attn_out_proj_weight__int8_scales": 485, + "p_model_model_language_model_layers_18_linear_attn_in_proj_a_weight__int8_scales": 486, + "p_model_model_language_model_layers_18_linear_attn_in_proj_b_weight__int8_scales": 487, + "p_model_model_language_model_layers_18_linear_attn_in_proj_qkv_weight__int8_scales": 488, + "p_model_model_language_model_layers_18_linear_attn_in_proj_z_weight__int8_scales": 489, + "p_model_model_language_model_layers_18_linear_attn_out_proj_weight__int8_scales": 490, + "p_model_model_language_model_layers_19_self_attn_k_proj_weight__int8_scales": 491, + "p_model_model_language_model_layers_19_self_attn_o_proj_weight__int8_scales": 492, + "p_model_model_language_model_layers_19_self_attn_q_proj_weight__int8_scales": 493, + "p_model_model_language_model_layers_19_self_attn_v_proj_weight__int8_scales": 494, + "p_model_model_language_model_layers_1_linear_attn_in_proj_a_weight__int8_scales": 495, + "p_model_model_language_model_layers_1_linear_attn_in_proj_b_weight__int8_scales": 496, + "p_model_model_language_model_layers_1_linear_attn_in_proj_qkv_weight__int8_scales": 497, + "p_model_model_language_model_layers_1_linear_attn_in_proj_z_weight__int8_scales": 498, + "p_model_model_language_model_layers_1_linear_attn_out_proj_weight__int8_scales": 499, + "p_model_model_language_model_layers_20_linear_attn_in_proj_a_weight__int8_scales": 500, + "p_model_model_language_model_layers_20_linear_attn_in_proj_b_weight__int8_scales": 501, + "p_model_model_language_model_layers_20_linear_attn_in_proj_qkv_weight__int8_scales": 502, + "p_model_model_language_model_layers_20_linear_attn_in_proj_z_weight__int8_scales": 503, + "p_model_model_language_model_layers_20_linear_attn_out_proj_weight__int8_scales": 504, + "p_model_model_language_model_layers_21_linear_attn_in_proj_a_weight__int8_scales": 505, + "p_model_model_language_model_layers_21_linear_attn_in_proj_b_weight__int8_scales": 506, + "p_model_model_language_model_layers_21_linear_attn_in_proj_qkv_weight__int8_scales": 507, + "p_model_model_language_model_layers_21_linear_attn_in_proj_z_weight__int8_scales": 508, + "p_model_model_language_model_layers_21_linear_attn_out_proj_weight__int8_scales": 509, + "p_model_model_language_model_layers_22_linear_attn_in_proj_a_weight__int8_scales": 510, + "p_model_model_language_model_layers_22_linear_attn_in_proj_b_weight__int8_scales": 511, + "p_model_model_language_model_layers_22_linear_attn_in_proj_qkv_weight__int8_scales": 512, + "p_model_model_language_model_layers_22_linear_attn_in_proj_z_weight__int8_scales": 513, + "p_model_model_language_model_layers_22_linear_attn_out_proj_weight__int8_scales": 514, + "p_model_model_language_model_layers_23_self_attn_k_proj_weight__int8_scales": 515, + "p_model_model_language_model_layers_23_self_attn_o_proj_weight__int8_scales": 516, + "p_model_model_language_model_layers_23_self_attn_q_proj_weight__int8_scales": 517, + "p_model_model_language_model_layers_23_self_attn_v_proj_weight__int8_scales": 518, + "p_model_model_language_model_layers_2_linear_attn_in_proj_a_weight__int8_scales": 519, + "p_model_model_language_model_layers_2_linear_attn_in_proj_b_weight__int8_scales": 520, + "p_model_model_language_model_layers_2_linear_attn_in_proj_qkv_weight__int8_scales": 521, + "p_model_model_language_model_layers_2_linear_attn_in_proj_z_weight__int8_scales": 522, + "p_model_model_language_model_layers_2_linear_attn_out_proj_weight__int8_scales": 523, + "p_model_model_language_model_layers_3_self_attn_k_proj_weight__int8_scales": 524, + "p_model_model_language_model_layers_3_self_attn_o_proj_weight__int8_scales": 525, + "p_model_model_language_model_layers_3_self_attn_q_proj_weight__int8_scales": 526, + "p_model_model_language_model_layers_3_self_attn_v_proj_weight__int8_scales": 527, + "p_model_model_language_model_layers_4_linear_attn_in_proj_a_weight__int8_scales": 528, + "p_model_model_language_model_layers_4_linear_attn_in_proj_b_weight__int8_scales": 529, + "p_model_model_language_model_layers_4_linear_attn_in_proj_qkv_weight__int8_scales": 530, + "p_model_model_language_model_layers_4_linear_attn_in_proj_z_weight__int8_scales": 531, + "p_model_model_language_model_layers_4_linear_attn_out_proj_weight__int8_scales": 532, + "p_model_model_language_model_layers_5_linear_attn_in_proj_a_weight__int8_scales": 533, + "p_model_model_language_model_layers_5_linear_attn_in_proj_b_weight__int8_scales": 534, + "p_model_model_language_model_layers_5_linear_attn_in_proj_qkv_weight__int8_scales": 535, + "p_model_model_language_model_layers_5_linear_attn_in_proj_z_weight__int8_scales": 536, + "p_model_model_language_model_layers_5_linear_attn_out_proj_weight__int8_scales": 537, + "p_model_model_language_model_layers_6_linear_attn_in_proj_a_weight__int8_scales": 538, + "p_model_model_language_model_layers_6_linear_attn_in_proj_b_weight__int8_scales": 539, + "p_model_model_language_model_layers_6_linear_attn_in_proj_qkv_weight__int8_scales": 540, + "p_model_model_language_model_layers_6_linear_attn_in_proj_z_weight__int8_scales": 541, + "p_model_model_language_model_layers_6_linear_attn_out_proj_weight__int8_scales": 542, + "p_model_model_language_model_layers_7_self_attn_k_proj_weight__int8_scales": 543, + "p_model_model_language_model_layers_7_self_attn_o_proj_weight__int8_scales": 544, + "p_model_model_language_model_layers_7_self_attn_q_proj_weight__int8_scales": 545, + "p_model_model_language_model_layers_7_self_attn_v_proj_weight__int8_scales": 546, + "p_model_model_language_model_layers_8_linear_attn_in_proj_a_weight__int8_scales": 547, + "p_model_model_language_model_layers_8_linear_attn_in_proj_b_weight__int8_scales": 548, + "p_model_model_language_model_layers_8_linear_attn_in_proj_qkv_weight__int8_scales": 549, + "p_model_model_language_model_layers_8_linear_attn_in_proj_z_weight__int8_scales": 550, + "p_model_model_language_model_layers_8_linear_attn_out_proj_weight__int8_scales": 551, + "p_model_model_language_model_layers_9_linear_attn_in_proj_a_weight__int8_scales": 552, + "p_model_model_language_model_layers_9_linear_attn_in_proj_b_weight__int8_scales": 553, + "p_model_model_language_model_layers_9_linear_attn_in_proj_qkv_weight__int8_scales": 554, + "p_model_model_language_model_layers_9_linear_attn_in_proj_z_weight__int8_scales": 555, + "p_model_model_language_model_layers_9_linear_attn_out_proj_weight__int8_scales": 556, + "p_model_model_language_model_embed_tokens_weight__rows_0_131072": 557, + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_0_131072": 558, + "p_model_model_language_model_embed_tokens_weight__rows_131072_117248": 559, + "p_model_model_language_model_embed_tokens_weight__int8_scales__rows_131072_117248": 560, + "embedding__part_0": 561, + "embedding__part_1": 562, + "embedding": 563, + "unsqueeze": 561, + "expand": 562, + "slice_1": 561, + "expand_1": 562, + "_to_copy": 564, + "matmul": 561, + "transpose": 562, + "select": 565, + "select_1": 566, + "slice_2": 562, + "slice_3": 566, + "copy": 565, + "transpose_1": 562, + "select_2": 561, + "slice_scatter": 567, + "select_scatter": 561, + "transpose_2": 562, + "transpose_7": 561, + "select_7": 565, + "slice_7": 567, + "transpose_8": 561, + "select_8": 565, + "slice_8": 561, + "copy_1": 565, + "transpose_9": 561, + "select_9": 562, + "slice_scatter_1": 568, + "select_scatter_1": 562, + "transpose_10": 561, + "transpose_12": 562, + "select_11": 565, + "cat": 561, + "cos": 562, + "mul": 569, + "sin": 562, + "mul_1": 561, + "mul_3": 562, + "linear": 570, + "transpose_13": 571, + "linear_1": 570, + "linear_2": 568, + "linear_3": 565, + "cat_1": 572, + "slice_10": 323, + "conv1d": 573, + "slice_11": 572, + "silu": 573, + "slice_12": 571, + "transpose_14": 572, + "slice_13": 571, + "slice_14": 573, + "slice_15": 574, + "sigmoid": 562, + "exp": 564, + "neg": 568, + "add_2": 572, + "softplus": 565, + "mul_4": 572, + "mul_5": 575, + "sum_1": 568, + "add_3": 565, + "rsqrt_1": 568, + "mul_6": 575, + "mul_7": 571, + "sum_2": 565, + "add_4": 568, + "rsqrt_2": 565, + "mul_8": 571, + "normalized_gated_delta_rule": 573, + "slice_16": 324, + "slice_17": 575, + "transpose_15": 574, + "mul_10": 575, + "mul_11": 574, + "linear_4": 562, + "add_6": 575, + "mul_13": 563, + "linear_5": 571, + "linear_6": 570, + "mul_14": 572, + "linear_7": 562, + "add_9": 563, + "mul_16": 562, + "linear_8": 571, + "transpose_16": 570, + "linear_9": 574, + "linear_10": 568, + "linear_11": 565, + "cat_2": 572, + "slice_18": 325, + "conv1d_1": 573, + "slice_19": 572, + "silu_3": 573, + "slice_20": 571, + "transpose_17": 570, + "slice_21": 575, + "slice_22": 571, + "slice_23": 572, + "sigmoid_1": 562, + "exp_1": 564, + "neg_1": 568, + "add_12": 570, + "softplus_1": 565, + "mul_17": 570, + "mul_18": 573, + "sum_3": 568, + "add_13": 565, + "rsqrt_6": 568, + "mul_19": 573, + "mul_20": 575, + "sum_4": 565, + "add_14": 568, + "rsqrt_7": 565, + "mul_21": 575, + "normalized_gated_delta_rule_1": 571, + "slice_24": 326, + "slice_25": 575, + "transpose_18": 570, + "mul_23": 575, + "mul_24": 571, + "linear_12": 562, + "add_16": 574, + "mul_26": 563, + "linear_13": 570, + "linear_14": 571, + "mul_27": 572, + "linear_15": 562, + "add_19": 563, + "mul_29": 562, + "linear_16": 570, + "transpose_19": 571, + "linear_17": 575, + "linear_18": 568, + "linear_19": 565, + "cat_3": 572, + "slice_26": 327, + "conv1d_2": 573, + "slice_27": 572, + "silu_6": 573, + "slice_28": 570, + "transpose_20": 571, + "slice_29": 574, + "slice_30": 570, + "slice_31": 572, + "sigmoid_2": 562, + "exp_2": 564, + "neg_2": 568, + "add_22": 571, + "softplus_2": 565, + "mul_30": 571, + "mul_31": 573, + "sum_5": 568, + "add_23": 565, + "rsqrt_11": 568, + "mul_32": 573, + "mul_33": 574, + "sum_6": 565, + "add_24": 568, + "rsqrt_12": 565, + "mul_34": 574, + "normalized_gated_delta_rule_2": 570, + "slice_32": 328, + "slice_33": 574, + "transpose_21": 571, + "mul_36": 574, + "mul_37": 570, + "linear_20": 562, + "add_26": 575, + "mul_39": 563, + "linear_21": 571, + "linear_22": 570, + "mul_40": 572, + "linear_23": 562, + "add_29": 563, + "mul_42": 562, + "linear_24": 571, + "slice_34": 574, + "slice_35": 575, + "mul_44": 570, + "transpose_22": 574, + "linear_25": 571, + "mul_46": 570, + "transpose_23": 571, + "linear_26": 570, + "transpose_24": 562, + "slice_36": 570, + "slice_37": 572, + "slice_38": 574, + "slice_39": 573, + "mul_47": 571, + "slice_40": 576, + "slice_41": 577, + "neg_3": 570, + "cat_4": 578, + "mul_48": 570, + "add_36": 578, + "mul_49": 577, + "slice_42": 576, + "slice_43": 571, + "neg_4": 574, + "cat_5": 570, + "mul_50": 576, + "add_37": 574, + "cat_6": 571, + "cat_7": 578, + "cache_update": 329, + "cache_update_1": 330, + "matmul_2": 574, + "transpose_26": 570, + "sigmoid_3": 574, + "mul_52": 575, + "linear_27": 562, + "add_39": 574, + "mul_54": 563, + "linear_28": 571, + "linear_29": 570, + "mul_55": 572, + "linear_30": 562, + "add_42": 563, + "mul_57": 562, + "linear_31": 571, + "transpose_27": 570, + "linear_32": 575, + "linear_33": 568, + "linear_34": 565, + "cat_8": 572, + "slice_44": 331, + "conv1d_3": 573, + "slice_45": 572, + "silu_10": 573, + "slice_46": 571, + "transpose_28": 570, + "slice_47": 574, + "slice_48": 571, + "slice_49": 572, + "sigmoid_4": 577, + "exp_3": 564, + "neg_5": 568, + "add_45": 576, + "softplus_3": 565, + "mul_58": 576, + "mul_59": 570, + "sum_7": 568, + "add_46": 565, + "rsqrt_20": 568, + "mul_60": 570, + "mul_61": 574, + "sum_8": 565, + "add_47": 568, + "rsqrt_21": 565, + "mul_62": 574, + "normalized_gated_delta_rule_3": 571, + "slice_50": 332, + "slice_51": 574, + "transpose_29": 570, + "mul_64": 574, + "mul_65": 571, + "linear_35": 562, + "add_49": 575, + "mul_67": 563, + "linear_36": 570, + "linear_37": 571, + "mul_68": 572, + "linear_38": 562, + "add_52": 563, + "mul_70": 562, + "linear_39": 570, + "transpose_30": 571, + "linear_40": 574, + "linear_41": 568, + "linear_42": 565, + "cat_9": 572, + "slice_52": 333, + "conv1d_4": 573, + "slice_53": 572, + "silu_13": 573, + "slice_54": 570, + "transpose_31": 571, + "slice_55": 575, + "slice_56": 570, + "slice_57": 572, + "sigmoid_5": 576, + "exp_4": 564, + "neg_6": 568, + "add_55": 577, + "softplus_4": 565, + "mul_71": 577, + "mul_72": 571, + "sum_9": 568, + "add_56": 565, + "rsqrt_25": 568, + "mul_73": 571, + "mul_74": 575, + "sum_10": 565, + "add_57": 568, + "rsqrt_26": 565, + "mul_75": 575, + "normalized_gated_delta_rule_4": 570, + "slice_58": 334, + "slice_59": 575, + "transpose_32": 571, + "mul_77": 575, + "mul_78": 570, + "linear_43": 562, + "add_59": 574, + "mul_80": 563, + "linear_44": 571, + "linear_45": 570, + "mul_81": 572, + "linear_46": 562, + "add_62": 563, + "mul_83": 562, + "linear_47": 571, + "transpose_33": 570, + "linear_48": 575, + "linear_49": 568, + "linear_50": 565, + "cat_10": 572, + "slice_60": 335, + "conv1d_5": 573, + "slice_61": 572, + "silu_16": 573, + "slice_62": 571, + "transpose_34": 570, + "slice_63": 574, + "slice_64": 571, + "slice_65": 572, + "sigmoid_6": 577, + "exp_5": 564, + "neg_7": 568, + "add_65": 576, + "softplus_5": 565, + "mul_84": 576, + "mul_85": 570, + "sum_11": 568, + "add_66": 565, + "rsqrt_30": 568, + "mul_86": 570, + "mul_87": 574, + "sum_12": 565, + "add_67": 568, + "rsqrt_31": 565, + "mul_88": 574, + "normalized_gated_delta_rule_5": 571, + "slice_66": 336, + "slice_67": 574, + "transpose_35": 570, + "mul_90": 574, + "mul_91": 571, + "linear_51": 562, + "add_69": 575, + "mul_93": 563, + "linear_52": 570, + "linear_53": 571, + "mul_94": 572, + "linear_54": 562, + "add_72": 563, + "mul_96": 562, + "linear_55": 570, + "slice_68": 574, + "slice_69": 575, + "mul_98": 571, + "transpose_36": 574, + "linear_56": 578, + "mul_100": 570, + "transpose_37": 578, + "linear_57": 571, + "transpose_38": 562, + "slice_70": 570, + "slice_71": 571, + "slice_72": 576, + "slice_73": 574, + "mul_101": 578, + "slice_74": 577, + "slice_75": 572, + "neg_8": 570, + "cat_11": 572, + "mul_102": 570, + "add_79": 572, + "mul_103": 577, + "slice_76": 578, + "slice_77": 570, + "neg_9": 576, + "cat_12": 570, + "mul_104": 576, + "add_80": 578, + "cat_13": 570, + "cat_14": 571, + "cache_update_2": 337, + "cache_update_3": 338, + "matmul_4": 574, + "transpose_40": 571, + "sigmoid_7": 574, + "mul_106": 575, + "linear_58": 562, + "add_82": 574, + "mul_108": 563, + "linear_59": 570, + "linear_60": 571, + "mul_109": 572, + "linear_61": 562, + "add_85": 563, + "mul_111": 562, + "linear_62": 570, + "transpose_41": 571, + "linear_63": 575, + "linear_64": 568, + "linear_65": 565, + "cat_15": 572, + "slice_78": 339, + "conv1d_6": 573, + "slice_79": 572, + "silu_20": 573, + "slice_80": 570, + "transpose_42": 571, + "slice_81": 574, + "slice_82": 570, + "slice_83": 572, + "sigmoid_8": 577, + "exp_6": 564, + "neg_10": 568, + "add_88": 576, + "softplus_6": 565, + "mul_112": 576, + "mul_113": 571, + "sum_13": 568, + "add_89": 565, + "rsqrt_39": 568, + "mul_114": 571, + "mul_115": 574, + "sum_14": 565, + "add_90": 568, + "rsqrt_40": 565, + "mul_116": 574, + "normalized_gated_delta_rule_6": 570, + "slice_84": 340, + "slice_85": 574, + "transpose_43": 571, + "mul_118": 574, + "mul_119": 570, + "linear_66": 562, + "add_92": 575, + "mul_121": 563, + "linear_67": 571, + "linear_68": 570, + "mul_122": 572, + "linear_69": 562, + "add_95": 563, + "mul_124": 562, + "linear_70": 571, + "transpose_44": 570, + "linear_71": 574, + "linear_72": 568, + "linear_73": 565, + "cat_16": 572, + "slice_86": 341, + "conv1d_7": 573, + "slice_87": 572, + "silu_23": 573, + "slice_88": 571, + "transpose_45": 570, + "slice_89": 575, + "slice_90": 571, + "slice_91": 572, + "sigmoid_9": 576, + "exp_7": 564, + "neg_11": 568, + "add_98": 577, + "softplus_7": 565, + "mul_125": 577, + "mul_126": 570, + "sum_15": 568, + "add_99": 565, + "rsqrt_44": 568, + "mul_127": 570, + "mul_128": 575, + "sum_16": 565, + "add_100": 568, + "rsqrt_45": 565, + "mul_129": 575, + "normalized_gated_delta_rule_7": 571, + "slice_92": 342, + "slice_93": 575, + "transpose_46": 570, + "mul_131": 575, + "mul_132": 571, + "linear_74": 562, + "add_102": 574, + "mul_134": 563, + "linear_75": 570, + "linear_76": 571, + "mul_135": 572, + "linear_77": 562, + "add_105": 563, + "mul_137": 562, + "linear_78": 570, + "transpose_47": 571, + "linear_79": 575, + "linear_80": 568, + "linear_81": 565, + "cat_17": 572, + "slice_94": 343, + "conv1d_8": 573, + "slice_95": 572, + "silu_26": 573, + "slice_96": 570, + "transpose_48": 571, + "slice_97": 574, + "slice_98": 570, + "slice_99": 572, + "sigmoid_10": 577, + "exp_8": 564, + "neg_12": 568, + "add_108": 576, + "softplus_8": 565, + "mul_138": 576, + "mul_139": 571, + "sum_17": 568, + "add_109": 565, + "rsqrt_49": 568, + "mul_140": 571, + "mul_141": 574, + "sum_18": 565, + "add_110": 568, + "rsqrt_50": 565, + "mul_142": 574, + "normalized_gated_delta_rule_8": 570, + "slice_100": 344, + "slice_101": 574, + "transpose_49": 571, + "mul_144": 574, + "mul_145": 570, + "linear_82": 562, + "add_112": 575, + "mul_147": 563, + "linear_83": 571, + "linear_84": 570, + "mul_148": 572, + "linear_85": 562, + "add_115": 563, + "mul_150": 562, + "linear_86": 571, + "slice_102": 574, + "slice_103": 575, + "mul_152": 570, + "transpose_50": 574, + "linear_87": 578, + "mul_154": 571, + "transpose_51": 578, + "linear_88": 570, + "transpose_52": 562, + "slice_104": 571, + "slice_105": 570, + "slice_106": 576, + "slice_107": 574, + "mul_155": 578, + "slice_108": 577, + "slice_109": 572, + "neg_13": 571, + "cat_18": 572, + "mul_156": 571, + "add_122": 572, + "mul_157": 577, + "slice_110": 578, + "slice_111": 571, + "neg_14": 576, + "cat_19": 571, + "mul_158": 576, + "add_123": 578, + "cat_20": 571, + "cat_21": 570, + "cache_update_4": 345, + "cache_update_5": 346, + "matmul_6": 574, + "transpose_54": 570, + "sigmoid_11": 574, + "mul_160": 575, + "linear_89": 562, + "add_125": 574, + "mul_162": 563, + "linear_90": 571, + "linear_91": 570, + "mul_163": 572, + "linear_92": 562, + "add_128": 563, + "mul_165": 562, + "linear_93": 571, + "transpose_55": 570, + "linear_94": 575, + "linear_95": 568, + "linear_96": 565, + "cat_22": 572, + "slice_112": 347, + "conv1d_9": 573, + "slice_113": 572, + "silu_30": 573, + "slice_114": 571, + "transpose_56": 570, + "slice_115": 574, + "slice_116": 571, + "slice_117": 572, + "sigmoid_12": 577, + "exp_9": 564, + "neg_15": 568, + "add_131": 576, + "softplus_9": 565, + "mul_166": 576, + "mul_167": 570, + "sum_19": 568, + "add_132": 565, + "rsqrt_58": 568, + "mul_168": 570, + "mul_169": 574, + "sum_20": 565, + "add_133": 568, + "rsqrt_59": 565, + "mul_170": 574, + "normalized_gated_delta_rule_9": 571, + "slice_118": 348, + "slice_119": 574, + "transpose_57": 570, + "mul_172": 574, + "mul_173": 571, + "linear_97": 562, + "add_135": 575, + "mul_175": 563, + "linear_98": 570, + "linear_99": 571, + "mul_176": 572, + "linear_100": 562, + "add_138": 563, + "mul_178": 562, + "linear_101": 570, + "transpose_58": 571, + "linear_102": 574, + "linear_103": 568, + "linear_104": 565, + "cat_23": 572, + "slice_120": 349, + "conv1d_10": 573, + "slice_121": 572, + "silu_33": 573, + "slice_122": 570, + "transpose_59": 571, + "slice_123": 575, + "slice_124": 570, + "slice_125": 572, + "sigmoid_13": 576, + "exp_10": 564, + "neg_16": 568, + "add_141": 577, + "softplus_10": 565, + "mul_179": 577, + "mul_180": 571, + "sum_21": 568, + "add_142": 565, + "rsqrt_63": 568, + "mul_181": 571, + "mul_182": 575, + "sum_22": 565, + "add_143": 568, + "rsqrt_64": 565, + "mul_183": 575, + "normalized_gated_delta_rule_10": 570, + "slice_126": 350, + "slice_127": 575, + "transpose_60": 571, + "mul_185": 575, + "mul_186": 570, + "linear_105": 562, + "add_145": 574, + "mul_188": 563, + "linear_106": 571, + "linear_107": 570, + "mul_189": 572, + "linear_108": 562, + "add_148": 563, + "mul_191": 562, + "linear_109": 571, + "transpose_61": 570, + "linear_110": 575, + "linear_111": 568, + "linear_112": 565, + "cat_24": 572, + "slice_128": 351, + "conv1d_11": 573, + "slice_129": 572, + "silu_36": 573, + "slice_130": 571, + "transpose_62": 570, + "slice_131": 574, + "slice_132": 571, + "slice_133": 572, + "sigmoid_14": 577, + "exp_11": 564, + "neg_17": 568, + "add_151": 576, + "softplus_11": 565, + "mul_192": 576, + "mul_193": 570, + "sum_23": 568, + "add_152": 565, + "rsqrt_68": 568, + "mul_194": 570, + "mul_195": 574, + "sum_24": 565, + "add_153": 568, + "rsqrt_69": 565, + "mul_196": 574, + "normalized_gated_delta_rule_11": 571, + "slice_134": 352, + "slice_135": 574, + "transpose_63": 570, + "mul_198": 574, + "mul_199": 571, + "linear_113": 562, + "add_155": 575, + "mul_201": 563, + "linear_114": 570, + "linear_115": 571, + "mul_202": 572, + "linear_116": 562, + "add_158": 563, + "mul_204": 562, + "linear_117": 570, + "slice_136": 574, + "slice_137": 575, + "mul_206": 571, + "transpose_64": 574, + "linear_118": 578, + "mul_208": 570, + "transpose_65": 578, + "linear_119": 571, + "transpose_66": 562, + "slice_138": 570, + "slice_139": 571, + "slice_140": 576, + "slice_141": 574, + "mul_209": 578, + "slice_142": 577, + "slice_143": 572, + "neg_18": 570, + "cat_25": 572, + "mul_210": 570, + "add_165": 572, + "mul_211": 577, + "slice_144": 578, + "slice_145": 570, + "neg_19": 576, + "cat_26": 570, + "mul_212": 576, + "add_166": 578, + "cat_27": 570, + "cat_28": 571, + "cache_update_6": 353, + "cache_update_7": 354, + "matmul_8": 574, + "transpose_68": 571, + "sigmoid_15": 574, + "mul_214": 575, + "linear_120": 562, + "add_168": 574, + "mul_216": 563, + "linear_121": 570, + "linear_122": 571, + "mul_217": 572, + "linear_123": 562, + "add_171": 563, + "mul_219": 562, + "linear_124": 570, + "transpose_69": 571, + "linear_125": 575, + "linear_126": 568, + "linear_127": 565, + "cat_29": 572, + "slice_146": 355, + "conv1d_12": 573, + "slice_147": 572, + "silu_40": 573, + "slice_148": 570, + "transpose_70": 571, + "slice_149": 574, + "slice_150": 570, + "slice_151": 572, + "sigmoid_16": 577, + "exp_12": 564, + "neg_20": 568, + "add_174": 576, + "softplus_12": 565, + "mul_220": 576, + "mul_221": 571, + "sum_25": 568, + "add_175": 565, + "rsqrt_77": 568, + "mul_222": 571, + "mul_223": 574, + "sum_26": 565, + "add_176": 568, + "rsqrt_78": 565, + "mul_224": 574, + "normalized_gated_delta_rule_12": 570, + "slice_152": 356, + "slice_153": 574, + "transpose_71": 571, + "mul_226": 574, + "mul_227": 570, + "linear_128": 562, + "add_178": 575, + "mul_229": 563, + "linear_129": 571, + "linear_130": 570, + "mul_230": 572, + "linear_131": 562, + "add_181": 563, + "mul_232": 562, + "linear_132": 571, + "transpose_72": 570, + "linear_133": 574, + "linear_134": 568, + "linear_135": 565, + "cat_30": 572, + "slice_154": 357, + "conv1d_13": 573, + "slice_155": 572, + "silu_43": 573, + "slice_156": 571, + "transpose_73": 570, + "slice_157": 575, + "slice_158": 571, + "slice_159": 572, + "sigmoid_17": 576, + "exp_13": 564, + "neg_21": 568, + "add_184": 577, + "softplus_13": 565, + "mul_233": 577, + "mul_234": 570, + "sum_27": 568, + "add_185": 565, + "rsqrt_82": 568, + "mul_235": 570, + "mul_236": 575, + "sum_28": 565, + "add_186": 568, + "rsqrt_83": 565, + "mul_237": 575, + "normalized_gated_delta_rule_13": 571, + "slice_160": 358, + "slice_161": 575, + "transpose_74": 570, + "mul_239": 575, + "mul_240": 571, + "linear_136": 562, + "add_188": 574, + "mul_242": 563, + "linear_137": 570, + "linear_138": 571, + "mul_243": 572, + "linear_139": 562, + "add_191": 563, + "mul_245": 562, + "linear_140": 570, + "transpose_75": 571, + "linear_141": 575, + "linear_142": 568, + "linear_143": 565, + "cat_31": 572, + "slice_162": 359, + "conv1d_14": 573, + "slice_163": 572, + "silu_46": 573, + "slice_164": 570, + "transpose_76": 571, + "slice_165": 574, + "slice_166": 570, + "slice_167": 572, + "sigmoid_18": 577, + "exp_14": 564, + "neg_22": 568, + "add_194": 576, + "softplus_14": 565, + "mul_246": 576, + "mul_247": 571, + "sum_29": 568, + "add_195": 565, + "rsqrt_87": 568, + "mul_248": 571, + "mul_249": 574, + "sum_30": 565, + "add_196": 568, + "rsqrt_88": 565, + "mul_250": 574, + "normalized_gated_delta_rule_14": 570, + "slice_168": 360, + "slice_169": 574, + "transpose_77": 571, + "mul_252": 574, + "mul_253": 570, + "linear_144": 562, + "add_198": 575, + "mul_255": 563, + "linear_145": 571, + "linear_146": 570, + "mul_256": 572, + "linear_147": 562, + "add_201": 563, + "mul_258": 562, + "linear_148": 571, + "slice_170": 574, + "slice_171": 575, + "mul_260": 570, + "transpose_78": 574, + "linear_149": 578, + "mul_262": 571, + "transpose_79": 578, + "linear_150": 570, + "transpose_80": 562, + "slice_172": 571, + "slice_173": 570, + "slice_174": 576, + "slice_175": 574, + "mul_263": 578, + "slice_176": 577, + "slice_177": 572, + "neg_23": 571, + "cat_32": 572, + "mul_264": 571, + "add_208": 572, + "mul_265": 577, + "slice_178": 578, + "slice_179": 571, + "neg_24": 576, + "cat_33": 571, + "mul_266": 576, + "add_209": 578, + "cat_34": 571, + "cat_35": 570, + "cache_update_8": 361, + "cache_update_9": 362, + "matmul_10": 574, + "transpose_82": 570, + "sigmoid_19": 574, + "mul_268": 575, + "linear_151": 562, + "add_211": 574, + "mul_270": 563, + "linear_152": 571, + "linear_153": 570, + "mul_271": 572, + "linear_154": 562, + "add_214": 563, + "mul_273": 562, + "linear_155": 571, + "transpose_83": 570, + "linear_156": 575, + "linear_157": 568, + "linear_158": 565, + "cat_36": 572, + "slice_180": 363, + "conv1d_15": 573, + "slice_181": 572, + "silu_50": 573, + "slice_182": 571, + "transpose_84": 570, + "slice_183": 574, + "slice_184": 571, + "slice_185": 572, + "sigmoid_20": 577, + "exp_15": 564, + "neg_25": 568, + "add_217": 576, + "softplus_15": 565, + "mul_274": 576, + "mul_275": 570, + "sum_31": 568, + "add_218": 565, + "rsqrt_96": 568, + "mul_276": 570, + "mul_277": 574, + "sum_32": 565, + "add_219": 568, + "rsqrt_97": 565, + "mul_278": 574, + "normalized_gated_delta_rule_15": 571, + "slice_186": 364, + "slice_187": 574, + "transpose_85": 570, + "mul_280": 574, + "mul_281": 571, + "linear_159": 562, + "add_221": 575, + "mul_283": 563, + "linear_160": 570, + "linear_161": 571, + "mul_284": 572, + "linear_162": 562, + "add_224": 563, + "mul_286": 562, + "linear_163": 570, + "transpose_86": 571, + "linear_164": 574, + "linear_165": 568, + "linear_166": 565, + "cat_37": 572, + "slice_188": 365, + "conv1d_16": 573, + "slice_189": 572, + "silu_53": 573, + "slice_190": 570, + "transpose_87": 571, + "slice_191": 575, + "slice_192": 570, + "slice_193": 572, + "sigmoid_21": 576, + "exp_16": 564, + "neg_26": 568, + "add_227": 577, + "softplus_16": 565, + "mul_287": 577, + "mul_288": 571, + "sum_33": 568, + "add_228": 565, + "rsqrt_101": 568, + "mul_289": 571, + "mul_290": 575, + "sum_34": 565, + "add_229": 568, + "rsqrt_102": 565, + "mul_291": 575, + "normalized_gated_delta_rule_16": 570, + "slice_194": 366, + "slice_195": 575, + "transpose_88": 571, + "mul_293": 575, + "mul_294": 570, + "linear_167": 562, + "add_231": 574, + "mul_296": 563, + "linear_168": 571, + "linear_169": 570, + "mul_297": 572, + "linear_170": 562, + "add_234": 563, + "mul_299": 562, + "linear_171": 571, + "transpose_89": 570, + "linear_172": 575, + "linear_173": 568, + "linear_174": 565, + "cat_38": 572, + "slice_196": 367, + "conv1d_17": 573, + "slice_197": 572, + "silu_56": 573, + "slice_198": 571, + "transpose_90": 570, + "slice_199": 574, + "slice_200": 571, + "slice_201": 572, + "sigmoid_22": 577, + "exp_17": 564, + "neg_27": 568, + "add_237": 576, + "softplus_17": 565, + "mul_300": 576, + "mul_301": 570, + "sum_35": 568, + "add_238": 565, + "rsqrt_106": 568, + "mul_302": 570, + "mul_303": 574, + "sum_36": 565, + "add_239": 568, + "rsqrt_107": 565, + "mul_304": 574, + "normalized_gated_delta_rule_17": 571, + "slice_202": 368, + "slice_203": 574, + "transpose_91": 570, + "mul_306": 574, + "mul_307": 571, + "linear_175": 562, + "add_241": 575, + "mul_309": 563, + "linear_176": 570, + "linear_177": 571, + "mul_310": 572, + "linear_178": 562, + "add_244": 563, + "mul_312": 562, + "linear_179": 570, + "slice_204": 574, + "slice_205": 575, + "mul_314": 571, + "transpose_92": 574, + "linear_180": 578, + "mul_316": 570, + "transpose_93": 578, + "linear_181": 571, + "transpose_94": 562, + "slice_206": 570, + "slice_207": 571, + "slice_208": 576, + "slice_209": 574, + "mul_317": 578, + "slice_210": 577, + "slice_211": 572, + "neg_28": 570, + "cat_39": 572, + "mul_318": 570, + "add_251": 572, + "mul_319": 577, + "slice_212": 569, + "slice_213": 578, + "neg_29": 576, + "cat_40": 578, + "mul_320": 576, + "add_252": 578, + "cat_41": 570, + "cat_42": 561, + "cache_update_10": 369, + "cache_update_11": 370, + "matmul_12": 574, + "transpose_96": 571, + "sigmoid_23": 574, + "mul_322": 575, + "linear_182": 561, + "add_254": 562, + "mul_324": 563, + "linear_183": 570, + "linear_184": 571, + "mul_325": 572, + "linear_185": 561, + "add_257": 563, + "mul_327": 562, + "linear_186__part_0": 579, + "linear_186__part_1": 580, + "linear_186": 581 + }, + "byteLength": 973959808 + }, + "rewrites": [ + { + "pattern": "rms_norm", + "nodes": [ + "pow_1", + "mean", + "add", + "rsqrt", + "mul_2", + "add_1", + "mul_3" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_2", + "mean_1", + "add_5", + "rsqrt_3", + "mul_9", + "mul_10" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_3", + "mean_2", + "add_7", + "rsqrt_4", + "mul_12", + "add_8", + "mul_13" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_4", + "mean_3", + "add_10", + "rsqrt_5", + "mul_15", + "add_11", + "mul_16" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_5", + "mean_4", + "add_15", + "rsqrt_8", + "mul_22", + "mul_23" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_6", + "mean_5", + "add_17", + "rsqrt_9", + "mul_25", + "add_18", + "mul_26" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_7", + "mean_6", + "add_20", + "rsqrt_10", + "mul_28", + "add_21", + "mul_29" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_8", + "mean_7", + "add_25", + "rsqrt_13", + "mul_35", + "mul_36" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_9", + "mean_8", + "add_27", + "rsqrt_14", + "mul_38", + "add_28", + "mul_39" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_10", + "mean_9", + "add_30", + "rsqrt_15", + "mul_41", + "add_31", + "mul_42" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_11", + "mean_10", + "add_32", + "rsqrt_16", + "mul_43", + "add_33", + "mul_44" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_12", + "mean_11", + "add_34", + "rsqrt_17", + "mul_45", + "add_35", + "mul_46" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_13", + "mean_12", + "add_40", + "rsqrt_18", + "mul_53", + "add_41", + "mul_54" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_14", + "mean_13", + "add_43", + "rsqrt_19", + "mul_56", + "add_44", + "mul_57" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_15", + "mean_14", + "add_48", + "rsqrt_22", + "mul_63", + "mul_64" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_16", + "mean_15", + "add_50", + "rsqrt_23", + "mul_66", + "add_51", + "mul_67" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_17", + "mean_16", + "add_53", + "rsqrt_24", + "mul_69", + "add_54", + "mul_70" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_18", + "mean_17", + "add_58", + "rsqrt_27", + "mul_76", + "mul_77" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_19", + "mean_18", + "add_60", + "rsqrt_28", + "mul_79", + "add_61", + "mul_80" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_20", + "mean_19", + "add_63", + "rsqrt_29", + "mul_82", + "add_64", + "mul_83" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_21", + "mean_20", + "add_68", + "rsqrt_32", + "mul_89", + "mul_90" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_22", + "mean_21", + "add_70", + "rsqrt_33", + "mul_92", + "add_71", + "mul_93" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_23", + "mean_22", + "add_73", + "rsqrt_34", + "mul_95", + "add_74", + "mul_96" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_24", + "mean_23", + "add_75", + "rsqrt_35", + "mul_97", + "add_76", + "mul_98" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_25", + "mean_24", + "add_77", + "rsqrt_36", + "mul_99", + "add_78", + "mul_100" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_26", + "mean_25", + "add_83", + "rsqrt_37", + "mul_107", + "add_84", + "mul_108" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_27", + "mean_26", + "add_86", + "rsqrt_38", + "mul_110", + "add_87", + "mul_111" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_28", + "mean_27", + "add_91", + "rsqrt_41", + "mul_117", + "mul_118" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_29", + "mean_28", + "add_93", + "rsqrt_42", + "mul_120", + "add_94", + "mul_121" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_30", + "mean_29", + "add_96", + "rsqrt_43", + "mul_123", + "add_97", + "mul_124" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_31", + "mean_30", + "add_101", + "rsqrt_46", + "mul_130", + "mul_131" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_32", + "mean_31", + "add_103", + "rsqrt_47", + "mul_133", + "add_104", + "mul_134" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_33", + "mean_32", + "add_106", + "rsqrt_48", + "mul_136", + "add_107", + "mul_137" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_34", + "mean_33", + "add_111", + "rsqrt_51", + "mul_143", + "mul_144" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_35", + "mean_34", + "add_113", + "rsqrt_52", + "mul_146", + "add_114", + "mul_147" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_36", + "mean_35", + "add_116", + "rsqrt_53", + "mul_149", + "add_117", + "mul_150" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_37", + "mean_36", + "add_118", + "rsqrt_54", + "mul_151", + "add_119", + "mul_152" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_38", + "mean_37", + "add_120", + "rsqrt_55", + "mul_153", + "add_121", + "mul_154" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_39", + "mean_38", + "add_126", + "rsqrt_56", + "mul_161", + "add_127", + "mul_162" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_40", + "mean_39", + "add_129", + "rsqrt_57", + "mul_164", + "add_130", + "mul_165" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_41", + "mean_40", + "add_134", + "rsqrt_60", + "mul_171", + "mul_172" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_42", + "mean_41", + "add_136", + "rsqrt_61", + "mul_174", + "add_137", + "mul_175" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_43", + "mean_42", + "add_139", + "rsqrt_62", + "mul_177", + "add_140", + "mul_178" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_44", + "mean_43", + "add_144", + "rsqrt_65", + "mul_184", + "mul_185" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_45", + "mean_44", + "add_146", + "rsqrt_66", + "mul_187", + "add_147", + "mul_188" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_46", + "mean_45", + "add_149", + "rsqrt_67", + "mul_190", + "add_150", + "mul_191" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_47", + "mean_46", + "add_154", + "rsqrt_70", + "mul_197", + "mul_198" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_48", + "mean_47", + "add_156", + "rsqrt_71", + "mul_200", + "add_157", + "mul_201" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_49", + "mean_48", + "add_159", + "rsqrt_72", + "mul_203", + "add_160", + "mul_204" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_50", + "mean_49", + "add_161", + "rsqrt_73", + "mul_205", + "add_162", + "mul_206" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_51", + "mean_50", + "add_163", + "rsqrt_74", + "mul_207", + "add_164", + "mul_208" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_52", + "mean_51", + "add_169", + "rsqrt_75", + "mul_215", + "add_170", + "mul_216" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_53", + "mean_52", + "add_172", + "rsqrt_76", + "mul_218", + "add_173", + "mul_219" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_54", + "mean_53", + "add_177", + "rsqrt_79", + "mul_225", + "mul_226" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_55", + "mean_54", + "add_179", + "rsqrt_80", + "mul_228", + "add_180", + "mul_229" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_56", + "mean_55", + "add_182", + "rsqrt_81", + "mul_231", + "add_183", + "mul_232" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_57", + "mean_56", + "add_187", + "rsqrt_84", + "mul_238", + "mul_239" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_58", + "mean_57", + "add_189", + "rsqrt_85", + "mul_241", + "add_190", + "mul_242" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_59", + "mean_58", + "add_192", + "rsqrt_86", + "mul_244", + "add_193", + "mul_245" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_60", + "mean_59", + "add_197", + "rsqrt_89", + "mul_251", + "mul_252" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_61", + "mean_60", + "add_199", + "rsqrt_90", + "mul_254", + "add_200", + "mul_255" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_62", + "mean_61", + "add_202", + "rsqrt_91", + "mul_257", + "add_203", + "mul_258" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_63", + "mean_62", + "add_204", + "rsqrt_92", + "mul_259", + "add_205", + "mul_260" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_64", + "mean_63", + "add_206", + "rsqrt_93", + "mul_261", + "add_207", + "mul_262" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_65", + "mean_64", + "add_212", + "rsqrt_94", + "mul_269", + "add_213", + "mul_270" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_66", + "mean_65", + "add_215", + "rsqrt_95", + "mul_272", + "add_216", + "mul_273" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_67", + "mean_66", + "add_220", + "rsqrt_98", + "mul_279", + "mul_280" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_68", + "mean_67", + "add_222", + "rsqrt_99", + "mul_282", + "add_223", + "mul_283" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_69", + "mean_68", + "add_225", + "rsqrt_100", + "mul_285", + "add_226", + "mul_286" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_70", + "mean_69", + "add_230", + "rsqrt_103", + "mul_292", + "mul_293" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_71", + "mean_70", + "add_232", + "rsqrt_104", + "mul_295", + "add_233", + "mul_296" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_72", + "mean_71", + "add_235", + "rsqrt_105", + "mul_298", + "add_236", + "mul_299" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_73", + "mean_72", + "add_240", + "rsqrt_108", + "mul_305", + "mul_306" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_74", + "mean_73", + "add_242", + "rsqrt_109", + "mul_308", + "add_243", + "mul_309" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_75", + "mean_74", + "add_245", + "rsqrt_110", + "mul_311", + "add_246", + "mul_312" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_76", + "mean_75", + "add_247", + "rsqrt_111", + "mul_313", + "add_248", + "mul_314" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_77", + "mean_76", + "add_249", + "rsqrt_112", + "mul_315", + "add_250", + "mul_316" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_78", + "mean_77", + "add_255", + "rsqrt_113", + "mul_323", + "add_256", + "mul_324" + ] + }, + { + "pattern": "rms_norm", + "nodes": [ + "pow_79", + "mean_78", + "add_258", + "rsqrt_114", + "mul_326", + "add_259", + "mul_327" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_1", + "mul_11" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_2", + "mul_14" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_4", + "mul_24" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_5", + "mul_27" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_7", + "mul_37" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_8", + "mul_40" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_9", + "mul_55" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_11", + "mul_65" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_12", + "mul_68" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_14", + "mul_78" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_15", + "mul_81" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_17", + "mul_91" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_18", + "mul_94" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_19", + "mul_109" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_21", + "mul_119" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_22", + "mul_122" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_24", + "mul_132" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_25", + "mul_135" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_27", + "mul_145" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_28", + "mul_148" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_29", + "mul_163" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_31", + "mul_173" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_32", + "mul_176" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_34", + "mul_186" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_35", + "mul_189" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_37", + "mul_199" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_38", + "mul_202" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_39", + "mul_217" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_41", + "mul_227" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_42", + "mul_230" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_44", + "mul_240" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_45", + "mul_243" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_47", + "mul_253" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_48", + "mul_256" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_49", + "mul_271" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_51", + "mul_281" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_52", + "mul_284" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_54", + "mul_294" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_55", + "mul_297" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_57", + "mul_307" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_58", + "mul_310" + ] + }, + { + "pattern": "silu_mul", + "nodes": [ + "silu_59", + "mul_325" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_1", + "clone_4", + "expand_2", + "unsqueeze_10", + "_unsafe_view_2", + "clone_5", + "expand_3", + "unsqueeze_11", + "transpose_25", + "matmul_1", + "mul_51", + "add_38", + "softmax", + "matmul_2" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_4", + "clone_11", + "expand_4", + "unsqueeze_17", + "_unsafe_view_5", + "clone_12", + "expand_5", + "unsqueeze_18", + "transpose_39", + "matmul_3", + "mul_105", + "add_81", + "softmax_1", + "matmul_4" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_7", + "clone_18", + "expand_6", + "unsqueeze_24", + "_unsafe_view_8", + "clone_19", + "expand_7", + "unsqueeze_25", + "transpose_53", + "matmul_5", + "mul_159", + "add_124", + "softmax_2", + "matmul_6" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_10", + "clone_25", + "expand_8", + "unsqueeze_31", + "_unsafe_view_11", + "clone_26", + "expand_9", + "unsqueeze_32", + "transpose_67", + "matmul_7", + "mul_213", + "add_167", + "softmax_3", + "matmul_8" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_13", + "clone_32", + "expand_10", + "unsqueeze_38", + "_unsafe_view_14", + "clone_33", + "expand_11", + "unsqueeze_39", + "transpose_81", + "matmul_9", + "mul_267", + "add_210", + "softmax_4", + "matmul_10" + ] + }, + { + "pattern": "attention", + "nodes": [ + "_unsafe_view_16", + "clone_39", + "expand_12", + "unsqueeze_45", + "_unsafe_view_17", + "clone_40", + "expand_13", + "unsqueeze_46", + "transpose_95", + "matmul_11", + "mul_321", + "add_253", + "softmax_5", + "matmul_12" + ] + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_1" + ], + "source": "b_model_model_language_model_rotary_emb_inv_freq" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_2" + ], + "source": "b_model_model_language_model_rotary_emb_inv_freq" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_3" + ], + "source": "b_model_model_language_model_rotary_emb_inv_freq" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_4" + ], + "source": "slice_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as" + ], + "source": "mul_3" + }, + { + "pattern": "elided_view", + "nodes": [ + "view" + ], + "source": "linear_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze" + ], + "source": "p_model_model_language_model_layers_0_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_5" + ], + "source": "p_model_model_language_model_layers_0_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_1" + ], + "source": "slice_13" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_2" + ], + "source": "slice_14" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_3" + ], + "source": "slice_15" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone" + ], + "source": "transpose_15" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_4" + ], + "source": "transpose_15" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_5" + ], + "source": "linear_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_6" + ], + "source": "mul_11" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_1" + ], + "source": "mul_13" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_2" + ], + "source": "mul_16" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_7" + ], + "source": "linear_9" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_1" + ], + "source": "p_model_model_language_model_layers_1_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_6" + ], + "source": "p_model_model_language_model_layers_1_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_8" + ], + "source": "slice_21" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_9" + ], + "source": "slice_22" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_10" + ], + "source": "slice_23" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_1" + ], + "source": "transpose_18" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_11" + ], + "source": "transpose_18" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_12" + ], + "source": "linear_9" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_13" + ], + "source": "mul_24" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_3" + ], + "source": "mul_26" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_4" + ], + "source": "mul_29" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_14" + ], + "source": "linear_17" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_2" + ], + "source": "p_model_model_language_model_layers_2_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_7" + ], + "source": "p_model_model_language_model_layers_2_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_15" + ], + "source": "slice_29" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_16" + ], + "source": "slice_30" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_17" + ], + "source": "slice_31" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_2" + ], + "source": "transpose_21" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_18" + ], + "source": "transpose_21" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_19" + ], + "source": "linear_17" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_20" + ], + "source": "mul_37" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_5" + ], + "source": "mul_39" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_6" + ], + "source": "mul_42" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_21" + ], + "source": "linear_24" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_3" + ], + "source": "slice_35" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view" + ], + "source": "slice_35" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_22" + ], + "source": "slice_34" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_7" + ], + "source": "mul_44" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_23" + ], + "source": "linear_25" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_8" + ], + "source": "mul_46" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_24" + ], + "source": "linear_26" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_8" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_9" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_6" + ], + "source": "transpose_26" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_25" + ], + "source": "transpose_26" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_9" + ], + "source": "mul_54" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_10" + ], + "source": "mul_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_26" + ], + "source": "linear_32" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_3" + ], + "source": "p_model_model_language_model_layers_4_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_12" + ], + "source": "p_model_model_language_model_layers_4_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_27" + ], + "source": "slice_47" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_28" + ], + "source": "slice_48" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_29" + ], + "source": "slice_49" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_7" + ], + "source": "transpose_29" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_30" + ], + "source": "transpose_29" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_31" + ], + "source": "linear_32" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_32" + ], + "source": "mul_65" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_11" + ], + "source": "mul_67" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_12" + ], + "source": "mul_70" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_33" + ], + "source": "linear_40" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_4" + ], + "source": "p_model_model_language_model_layers_5_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_13" + ], + "source": "p_model_model_language_model_layers_5_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_34" + ], + "source": "slice_55" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_35" + ], + "source": "slice_56" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_36" + ], + "source": "slice_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_8" + ], + "source": "transpose_32" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_37" + ], + "source": "transpose_32" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_38" + ], + "source": "linear_40" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_39" + ], + "source": "mul_78" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_13" + ], + "source": "mul_80" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_14" + ], + "source": "mul_83" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_40" + ], + "source": "linear_48" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_5" + ], + "source": "p_model_model_language_model_layers_6_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_14" + ], + "source": "p_model_model_language_model_layers_6_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_41" + ], + "source": "slice_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_42" + ], + "source": "slice_64" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_43" + ], + "source": "slice_65" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_9" + ], + "source": "transpose_35" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_44" + ], + "source": "transpose_35" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_45" + ], + "source": "linear_48" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_46" + ], + "source": "mul_91" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_15" + ], + "source": "mul_93" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_16" + ], + "source": "mul_96" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_47" + ], + "source": "linear_55" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_10" + ], + "source": "slice_69" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_3" + ], + "source": "slice_69" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_48" + ], + "source": "slice_68" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_17" + ], + "source": "mul_98" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_49" + ], + "source": "linear_56" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_18" + ], + "source": "mul_100" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_50" + ], + "source": "linear_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_15" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_16" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_13" + ], + "source": "transpose_40" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_51" + ], + "source": "transpose_40" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_19" + ], + "source": "mul_108" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_20" + ], + "source": "mul_111" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_52" + ], + "source": "linear_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_6" + ], + "source": "p_model_model_language_model_layers_8_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_19" + ], + "source": "p_model_model_language_model_layers_8_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_53" + ], + "source": "slice_81" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_54" + ], + "source": "slice_82" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_55" + ], + "source": "slice_83" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_14" + ], + "source": "transpose_43" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_56" + ], + "source": "transpose_43" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_57" + ], + "source": "linear_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_58" + ], + "source": "mul_119" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_21" + ], + "source": "mul_121" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_22" + ], + "source": "mul_124" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_59" + ], + "source": "linear_71" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_7" + ], + "source": "p_model_model_language_model_layers_9_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_20" + ], + "source": "p_model_model_language_model_layers_9_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_60" + ], + "source": "slice_89" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_61" + ], + "source": "slice_90" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_62" + ], + "source": "slice_91" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_15" + ], + "source": "transpose_46" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_63" + ], + "source": "transpose_46" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_64" + ], + "source": "linear_71" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_65" + ], + "source": "mul_132" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_23" + ], + "source": "mul_134" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_24" + ], + "source": "mul_137" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_66" + ], + "source": "linear_79" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_8" + ], + "source": "p_model_model_language_model_layers_10_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_21" + ], + "source": "p_model_model_language_model_layers_10_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_67" + ], + "source": "slice_97" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_68" + ], + "source": "slice_98" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_69" + ], + "source": "slice_99" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_16" + ], + "source": "transpose_49" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_70" + ], + "source": "transpose_49" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_71" + ], + "source": "linear_79" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_72" + ], + "source": "mul_145" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_25" + ], + "source": "mul_147" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_26" + ], + "source": "mul_150" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_73" + ], + "source": "linear_86" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_17" + ], + "source": "slice_103" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_6" + ], + "source": "slice_103" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_74" + ], + "source": "slice_102" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_27" + ], + "source": "mul_152" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_75" + ], + "source": "linear_87" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_28" + ], + "source": "mul_154" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_76" + ], + "source": "linear_88" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_22" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_23" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_20" + ], + "source": "transpose_54" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_77" + ], + "source": "transpose_54" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_29" + ], + "source": "mul_162" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_30" + ], + "source": "mul_165" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_78" + ], + "source": "linear_94" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_9" + ], + "source": "p_model_model_language_model_layers_12_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_26" + ], + "source": "p_model_model_language_model_layers_12_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_79" + ], + "source": "slice_115" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_80" + ], + "source": "slice_116" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_81" + ], + "source": "slice_117" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_21" + ], + "source": "transpose_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_82" + ], + "source": "transpose_57" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_83" + ], + "source": "linear_94" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_84" + ], + "source": "mul_173" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_31" + ], + "source": "mul_175" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_32" + ], + "source": "mul_178" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_85" + ], + "source": "linear_102" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_10" + ], + "source": "p_model_model_language_model_layers_13_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_27" + ], + "source": "p_model_model_language_model_layers_13_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_86" + ], + "source": "slice_123" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_87" + ], + "source": "slice_124" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_88" + ], + "source": "slice_125" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_22" + ], + "source": "transpose_60" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_89" + ], + "source": "transpose_60" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_90" + ], + "source": "linear_102" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_91" + ], + "source": "mul_186" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_33" + ], + "source": "mul_188" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_34" + ], + "source": "mul_191" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_92" + ], + "source": "linear_110" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_11" + ], + "source": "p_model_model_language_model_layers_14_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_28" + ], + "source": "p_model_model_language_model_layers_14_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_93" + ], + "source": "slice_131" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_94" + ], + "source": "slice_132" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_95" + ], + "source": "slice_133" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_23" + ], + "source": "transpose_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_96" + ], + "source": "transpose_63" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_97" + ], + "source": "linear_110" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_98" + ], + "source": "mul_199" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_35" + ], + "source": "mul_201" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_36" + ], + "source": "mul_204" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_99" + ], + "source": "linear_117" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_24" + ], + "source": "slice_137" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_9" + ], + "source": "slice_137" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_100" + ], + "source": "slice_136" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_37" + ], + "source": "mul_206" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_101" + ], + "source": "linear_118" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_38" + ], + "source": "mul_208" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_102" + ], + "source": "linear_119" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_29" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_30" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_27" + ], + "source": "transpose_68" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_103" + ], + "source": "transpose_68" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_39" + ], + "source": "mul_216" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_40" + ], + "source": "mul_219" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_104" + ], + "source": "linear_125" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_12" + ], + "source": "p_model_model_language_model_layers_16_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_33" + ], + "source": "p_model_model_language_model_layers_16_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_105" + ], + "source": "slice_149" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_106" + ], + "source": "slice_150" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_107" + ], + "source": "slice_151" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_28" + ], + "source": "transpose_71" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_108" + ], + "source": "transpose_71" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_109" + ], + "source": "linear_125" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_110" + ], + "source": "mul_227" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_41" + ], + "source": "mul_229" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_42" + ], + "source": "mul_232" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_111" + ], + "source": "linear_133" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_13" + ], + "source": "p_model_model_language_model_layers_17_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_34" + ], + "source": "p_model_model_language_model_layers_17_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_112" + ], + "source": "slice_157" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_113" + ], + "source": "slice_158" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_114" + ], + "source": "slice_159" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_29" + ], + "source": "transpose_74" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_115" + ], + "source": "transpose_74" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_116" + ], + "source": "linear_133" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_117" + ], + "source": "mul_240" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_43" + ], + "source": "mul_242" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_44" + ], + "source": "mul_245" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_118" + ], + "source": "linear_141" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_14" + ], + "source": "p_model_model_language_model_layers_18_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_35" + ], + "source": "p_model_model_language_model_layers_18_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_119" + ], + "source": "slice_165" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_120" + ], + "source": "slice_166" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_121" + ], + "source": "slice_167" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_30" + ], + "source": "transpose_77" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_122" + ], + "source": "transpose_77" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_123" + ], + "source": "linear_141" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_124" + ], + "source": "mul_253" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_45" + ], + "source": "mul_255" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_46" + ], + "source": "mul_258" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_125" + ], + "source": "linear_148" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_31" + ], + "source": "slice_171" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_12" + ], + "source": "slice_171" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_126" + ], + "source": "slice_170" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_47" + ], + "source": "mul_260" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_127" + ], + "source": "linear_149" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_48" + ], + "source": "mul_262" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_128" + ], + "source": "linear_150" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_36" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_37" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_34" + ], + "source": "transpose_82" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_129" + ], + "source": "transpose_82" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_49" + ], + "source": "mul_270" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_50" + ], + "source": "mul_273" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_130" + ], + "source": "linear_156" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_15" + ], + "source": "p_model_model_language_model_layers_20_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_40" + ], + "source": "p_model_model_language_model_layers_20_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_131" + ], + "source": "slice_183" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_132" + ], + "source": "slice_184" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_133" + ], + "source": "slice_185" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_35" + ], + "source": "transpose_85" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_134" + ], + "source": "transpose_85" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_135" + ], + "source": "linear_156" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_136" + ], + "source": "mul_281" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_51" + ], + "source": "mul_283" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_52" + ], + "source": "mul_286" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_137" + ], + "source": "linear_164" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_16" + ], + "source": "p_model_model_language_model_layers_21_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_41" + ], + "source": "p_model_model_language_model_layers_21_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_138" + ], + "source": "slice_191" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_139" + ], + "source": "slice_192" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_140" + ], + "source": "slice_193" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_36" + ], + "source": "transpose_88" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_141" + ], + "source": "transpose_88" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_142" + ], + "source": "linear_164" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_143" + ], + "source": "mul_294" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_53" + ], + "source": "mul_296" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_54" + ], + "source": "mul_299" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_144" + ], + "source": "linear_172" + }, + { + "pattern": "elided_view", + "nodes": [ + "squeeze_17" + ], + "source": "p_model_model_language_model_layers_22_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_42" + ], + "source": "p_model_model_language_model_layers_22_linear_attn_conv1d_weight" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_145" + ], + "source": "slice_199" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_146" + ], + "source": "slice_200" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_147" + ], + "source": "slice_201" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_37" + ], + "source": "transpose_91" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_148" + ], + "source": "transpose_91" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_149" + ], + "source": "linear_172" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_150" + ], + "source": "mul_307" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_55" + ], + "source": "mul_309" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_56" + ], + "source": "mul_312" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_151" + ], + "source": "linear_179" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_38" + ], + "source": "slice_205" + }, + { + "pattern": "elided_view", + "nodes": [ + "_unsafe_view_15" + ], + "source": "slice_205" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_152" + ], + "source": "slice_204" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_57" + ], + "source": "mul_314" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_153" + ], + "source": "linear_180" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_58" + ], + "source": "mul_316" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_154" + ], + "source": "linear_181" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_43" + ], + "source": "mul" + }, + { + "pattern": "elided_view", + "nodes": [ + "unsqueeze_44" + ], + "source": "mul_1" + }, + { + "pattern": "elided_view", + "nodes": [ + "clone_41" + ], + "source": "transpose_96" + }, + { + "pattern": "elided_view", + "nodes": [ + "view_155" + ], + "source": "transpose_96" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_59" + ], + "source": "mul_324" + }, + { + "pattern": "elided_view", + "nodes": [ + "type_as_60" + ], + "source": "mul_327" + }, + { + "pattern": "elided_view", + "nodes": [ + "alias" + ], + "source": "mul_327" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_10" + ], + "input": "states_0", + "output": "slice_10", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_16" + ], + "input": "states_1", + "output": "slice_16", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_18" + ], + "input": "states_2", + "output": "slice_18", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_24" + ], + "input": "states_3", + "output": "slice_24", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_26" + ], + "input": "states_4", + "output": "slice_26", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_32" + ], + "input": "states_5", + "output": "slice_32", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_44" + ], + "input": "states_8", + "output": "slice_44", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_50" + ], + "input": "states_9", + "output": "slice_50", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_52" + ], + "input": "states_10", + "output": "slice_52", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_58" + ], + "input": "states_11", + "output": "slice_58", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_60" + ], + "input": "states_12", + "output": "slice_60", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_66" + ], + "input": "states_13", + "output": "slice_66", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_78" + ], + "input": "states_16", + "output": "slice_78", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_84" + ], + "input": "states_17", + "output": "slice_84", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_86" + ], + "input": "states_18", + "output": "slice_86", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_92" + ], + "input": "states_19", + "output": "slice_92", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_94" + ], + "input": "states_20", + "output": "slice_94", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_100" + ], + "input": "states_21", + "output": "slice_100", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_112" + ], + "input": "states_24", + "output": "slice_112", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_118" + ], + "input": "states_25", + "output": "slice_118", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_120" + ], + "input": "states_26", + "output": "slice_120", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_126" + ], + "input": "states_27", + "output": "slice_126", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_128" + ], + "input": "states_28", + "output": "slice_128", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_134" + ], + "input": "states_29", + "output": "slice_134", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_146" + ], + "input": "states_32", + "output": "slice_146", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_152" + ], + "input": "states_33", + "output": "slice_152", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_154" + ], + "input": "states_34", + "output": "slice_154", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_160" + ], + "input": "states_35", + "output": "slice_160", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_162" + ], + "input": "states_36", + "output": "slice_162", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_168" + ], + "input": "states_37", + "output": "slice_168", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_180" + ], + "input": "states_40", + "output": "slice_180", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_186" + ], + "input": "states_41", + "output": "slice_186", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_188" + ], + "input": "states_42", + "output": "slice_188", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_194" + ], + "input": "states_43", + "output": "slice_194", + "kind": "recurrent" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_196" + ], + "input": "states_44", + "output": "slice_196", + "kind": "convolution" + }, + { + "pattern": "dead_state_output_alias", + "nodes": [ + "slice_202" + ], + "input": "states_45", + "output": "slice_202", + "kind": "recurrent" + } + ], + "partitions": [ + { + "node": "embedding", + "weight": "p_model_model_language_model_embed_tokens_weight", + "rowsPerShard": 131072, + "bindingLimitBytes": 134217728 + }, + { + "node": "linear_186", + "weight": "p_model_model_language_model_embed_tokens_weight", + "rowsPerShard": 131072, + "bindingLimitBytes": 134217728 + } + ] +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/graph.json b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/graph.json new file mode 100644 index 0000000000000000000000000000000000000000..bdffc0d2cee393d9feae3209f1281af3a4d4c3c1 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/graph.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:32d7ed995ec990c9ef6751dd8c2dd5824068d5816a4c070f43b8f8f74bdb59fe +size 10602779 diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d784bb44e0ee0b81f94ebb681a458b9de53a00dc --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/004c06d76ce77979b63fe4450df54986355fd10cd2ef62adae86780e49e4a4a7.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 16384u; +if (i >= 16384u) { return; } +out[i] = f32(f32(b0[i]) + (f32(b1[i]) * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a0d5268212599675490c8bf45f22a240ac7aeb17 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0112d93bb1aa347d8887ce1b95610a1a21fc8450ff8f59bcd37747fa6bf6fa52.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +out[i] = f32(f32(b0[i]) + (f32(b1[i]) * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3685c38c999329a75de1b14ae10134b4a83b8ff8 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0151a813c4a92adeb44f0e3b6108c77b29804118b588c9025805f724fce839b4.wgsl @@ -0,0 +1,24 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var factor: f32; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.x + group.y * 128u; + if (row >= 128u) { return; } + let lane = local.x; + if (lane == 0u) { + var total = 0.0; + for (var j = 0u; j < 256u; j++) { + let v = f32(b0[row * 256u + j]); + total += v * v; + } + factor = inverseSqrt(total / 256.0 + 1e-06); + } + workgroupBarrier(); + for (var p = lane; p < 256u; p += 64u) { + let i = row * 256u + p; + out[i] = f32(f32(b0[row * 256u + p]) * factor * (f32(b1[p]) + 1.0)); + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0153b6b641515b31ea2980613978c809e225435001b48fbec69afa520251a7af.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0153b6b641515b31ea2980613978c809e225435001b48fbec69afa520251a7af.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9e81e5f1a534b5920d13fc85a4235bc7494846d3 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0153b6b641515b31ea2980613978c809e225435001b48fbec69afa520251a7af.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 16>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 16u; + var acc: array, 1>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 1u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 4u && col < 1024u) { value = f32(f32(b0[(batch * 4u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 117248u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 1u; r++) { + let a_value = tile_a[local.y * 1u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 1u; r++) { + let out_row = tile_row + local.y * 1u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 4u && out_col < 117248u) { + let i = (batch * 4u + out_row) * 117248u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/026b53c4ecb0527a45fbd6020a725ad4d0ea2ea28c8cfa0a56e19cb0c1ad08ea.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/026b53c4ecb0527a45fbd6020a725ad4d0ea2ea28c8cfa0a56e19cb0c1ad08ea.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0311e706bf36190adc9732d32a07054e6cc23bbe --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/026b53c4ecb0527a45fbd6020a725ad4d0ea2ea28c8cfa0a56e19cb0c1ad08ea.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 3973120u; +if (i >= 3973120u) { return; } +let coord = (i / 1u) % 248320u; +if (coord >= 0u && coord < 131072u) { out[i] = f32(b0[(i / 248320u) * 131072u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 131072u && coord < 248320u) { out[i] = f32(b1[(i / 248320u) * 117248u + (coord - 131072u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0337d8a473b6e485bd383533df7d373dbbfe7a5f3a982424004560b7213fbaa3.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0337d8a473b6e485bd383533df7d373dbbfe7a5f3a982424004560b7213fbaa3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c50a7f2acd01b33468240fbe94eab2c802c81995 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0337d8a473b6e485bd383533df7d373dbbfe7a5f3a982424004560b7213fbaa3.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 4096u; +if (i >= 4096u) { return; } +let coord = (i / 1u) % 64u; +if (coord >= 0u && coord < 32u) { out[i] = f32(b0[(i / 64u) * 32u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 32u && coord < 64u) { out[i] = f32(b0[(i / 64u) * 32u + (coord - 32u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/04f26661cc9ac73bb7f39f66b76c4be31ee8774e094eeb41ed7d677e1b1d3fa9.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/04f26661cc9ac73bb7f39f66b76c4be31ee8774e094eeb41ed7d677e1b1d3fa9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..67133bd5fb075efe996a921acde0d02717ffa6f3 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/04f26661cc9ac73bb7f39f66b76c4be31ee8774e094eeb41ed7d677e1b1d3fa9.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 10u) { return; } +out[i] = f32(b0[((i / 10u) % 1u) * 32u + ((i / 10u) % 1u) * 32u + (((i / 1u) % 10u) * 3u + 2u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2c6541fe94d64690204bfe16ca3317d9d1fb5108 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/04f5cb70141e507d48bd44a76bfdf264a404a706e8a012502dce5a22ee75bab0.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 24576u; +if (i >= 24576u) { return; } +out[i] = f32(b0[((i / 24576u) % 1u) * 24576u + ((i / 4u) % 6144u) * 1u + ((i / 1u) % 4u) * 6144u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..587dbe5e99d52f935166db2e2b5f2e8af36c39cb --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/053386c9a33cc1c62378ebe3b3614c6b9ee662d6004a1f27bc616653e62cc24b.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 64>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 64u; + var acc: array, 4>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 4u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 64u && col < 1024u) { value = f32(f32(b0[(batch * 64u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 2048u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 4u; r++) { + let a_value = tile_a[local.y * 4u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 4u; r++) { + let out_row = tile_row + local.y * 4u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 64u && out_col < 2048u) { + let i = (batch * 64u + out_row) * 2048u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0554a443ecf6ead32fd2d58178b9ce3c592c89a09b3f63fa509ee947323bb497.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0554a443ecf6ead32fd2d58178b9ce3c592c89a09b3f63fa509ee947323bb497.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..586ebe23e7ab6f7d0ea3d4db2ae21148fc2ede1a --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0554a443ecf6ead32fd2d58178b9ce3c592c89a09b3f63fa509ee947323bb497.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 128u; +if (i >= 128u) { return; } +out[i] = f32(b0[((i / 128u) % 1u) * 512u + ((i / 64u) % 2u) * 256u + ((i / 64u) % 1u) * 256u + (((i / 1u) % 64u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/06eecde176beab9ff488ad533f2f4513e09bc49a30c4358f9648076f619f9992.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/06eecde176beab9ff488ad533f2f4513e09bc49a30c4358f9648076f619f9992.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ee8c8a4cc0cdaea931261f0dc33dc98926ba77b5 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/06eecde176beab9ff488ad533f2f4513e09bc49a30c4358f9648076f619f9992.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 6144u; +if (i >= 6144u) { return; } +out[i] = f32(b0[((i / 2048u) % 3u) * 2048u + ((i / 2048u) % 1u) * 2048u + ((i / 64u) % 32u) * 1u + ((i / 1u) % 64u) * 32u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0725908abf78c022f86deb338a06779f6f1346ebfe54ff5a43ee6b9eeb93c972.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0725908abf78c022f86deb338a06779f6f1346ebfe54ff5a43ee6b9eeb93c972.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..71b292e22afb334a345d95567bc425387a266f55 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0725908abf78c022f86deb338a06779f6f1346ebfe54ff5a43ee6b9eeb93c972.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 44u) { return; } +out[i] = f32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..aae8351dd54da00a71ffebf7e3c9ee6e67fcbfde --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/07cfabe2e2c2feff22a06c8296600e5ca21efb7444419f0a05e5ff6f48e0292d.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +let x = f32(b0[i]); +out[i] = f32(-x); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0835fa80eaadb7c0e6d677e1fc622ad5dcedabb431e5d312f3b7aeb2252a7c7c.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0835fa80eaadb7c0e6d677e1fc622ad5dcedabb431e5d312f3b7aeb2252a7c7c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..91f740aa7006d14dca451533719861c5823ff016 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0835fa80eaadb7c0e6d677e1fc622ad5dcedabb431e5d312f3b7aeb2252a7c7c.wgsl @@ -0,0 +1,8 @@ +@group(0) @binding(0) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +out[i] = f32(0.0); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d671cc16305ddbecbb23cd39809cff43b1852827 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/08390b1553d11820d1d96b10add657b92acb8d26cd9f11acab8f16777ecfc10a.wgsl @@ -0,0 +1,21 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 436224u; +if (i >= 436224u) { return; } +let batch = i / 436224u; +let channel = (i / 71u) % 6144u; +let group = channel / 1u; +var acc = 0.0; +for (var c = 0u; c < 1u; c++) { + for (var p = 0u; p < 4u; p++) { + let pos0 = i32((i / 1u) % 71u) * 1 + i32((p / 1u) % 4u) * 1 - 3; + if (pos0 >= 0 && pos0 < 68) { acc += f32(b0[batch * 417792u + (group * 1u + c) * 68u + u32(pos0) * 1u]) * f32(b1[(channel * 1u + c) * 4u + p]); } + } +} +out[i] = f32(acc + 0.0); + +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..49a08ac343cccf84ad9ff97c0e9afb2e3f61c0a4 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/09ea34c2533b929bd309e2bcbcf6d27fa02c731d82276ba6425db85961ea121b.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 256u; +if (i >= 256u) { return; } +let x = f32(b0[i]); +out[i] = f32(inverseSqrt(x)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0b8815b1efe176d55ab08530c0a9e1913cf6c9ef3e773c574843573d237cf196.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0b8815b1efe176d55ab08530c0a9e1913cf6c9ef3e773c574843573d237cf196.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..686ab91f45e66184b49d0f2e7429a2171e634bad --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0b8815b1efe176d55ab08530c0a9e1913cf6c9ef3e773c574843573d237cf196.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 256u; +if (i >= 256u) { return; } +let x = f32(b0[i]); +out[i] = f32(sin(x)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..54d385f25c67cd568fe09262762d9a1904e683d6 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0bbd6a5e356c1205a61dc3580a3ff68d88b98a952a5692d66b56ded171d9fc56.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +out[i] = f32(f32(b0[i]) + (1e-06 * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0d1edb63bf4dc212d55f8a4c79a50e04c0d822bd9f6e193003e2071ba0d57d69.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0d1edb63bf4dc212d55f8a4c79a50e04c0d822bd9f6e193003e2071ba0d57d69.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..29b5fc50bb132d2ebad8fbc34a147c39968c70ec --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0d1edb63bf4dc212d55f8a4c79a50e04c0d822bd9f6e193003e2071ba0d57d69.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +out[i] = f32(b0[((i / 8192u) % 1u) * 16384u + ((i / 2048u) % 4u) * 4096u + ((i / 256u) % 8u) * 512u + (((i / 1u) % 256u) * 1u + 256u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0d3892883d1532d173a7dbe85d9ab1daa87f0d9b4923a535b8d2003f24f9500c.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0d3892883d1532d173a7dbe85d9ab1daa87f0d9b4923a535b8d2003f24f9500c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8520366c6db35027de02453b6c726e0fc678ab4f --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0d3892883d1532d173a7dbe85d9ab1daa87f0d9b4923a535b8d2003f24f9500c.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 256u; +if (i >= 256u) { return; } +let coord = (i / 1u) % 64u; +if (coord >= 0u && coord < 32u) { out[i] = f32(b0[(i / 64u) * 32u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 32u && coord < 64u) { out[i] = f32(b0[(i / 64u) * 32u + (coord - 32u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..6d67f208ab0873a37709d785f6d09c3ac0ef60de --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0ef8fd3c2fe267549e7de4eb9a074d2ada8e80c91ce2f9c9c4d6b24d17742821.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 2048u; + let col = index % 2048u; + let word = b1[row * 512u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 64u + col / 32u]; +} + +var tile_a: array, 64>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 64u; + var acc: array, 4>; + for (var k0 = 0u; k0 < 2048u; k0 += 16u) { + for (var e = 0u; e < 4u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 64u && col < 2048u) { value = f32(f32(b0[(batch * 64u + row) * 2048u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 1024u && first < 2048u) { + word = b1[n_index * 512u + first / 4u]; + scale = f32(b2[n_index * 64u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 2048u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 4u; r++) { + let a_value = tile_a[local.y * 4u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 4u; r++) { + let out_row = tile_row + local.y * 4u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 64u && out_col < 1024u) { + let i = (batch * 64u + out_row) * 1024u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0f943f386d2f37a8e56b4285b26e1c093c2f2fa2a5ce81cc43c44cc9bb0137f5.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0f943f386d2f37a8e56b4285b26e1c093c2f2fa2a5ce81cc43c44cc9bb0137f5.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..15c550b1f32e4f6bb9a861ae384749b50592bd5d --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0f943f386d2f37a8e56b4285b26e1c093c2f2fa2a5ce81cc43c44cc9bb0137f5.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 384u; +if (i >= 384u) { return; } +out[i] = f32(b0[((i / 128u) % 3u) * 128u + ((i / 128u) % 1u) * 128u + ((i / 32u) % 4u) * 1u + ((i / 1u) % 32u) * 4u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..43d5bb662e9175428e6b81bc2f0e2980727497a7 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/0fbc9484ef8ca703ced4ffd123f1ae733c44556b6cb160ee803b820ff5ec867d.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +out[i] = f32(f32(b0[i]) * f32(b1[((i / 64u) % 16u) * 64u + ((i / 1u) % 64u) * 1u])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/11a92e9cba6bfc381c98c5127db137639d6c87f1cb220459880ead039dc81256.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/11a92e9cba6bfc381c98c5127db137639d6c87f1cb220459880ead039dc81256.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..34af207bc7ce034721ed448af19bbd095c4bdb49 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/11a92e9cba6bfc381c98c5127db137639d6c87f1cb220459880ead039dc81256.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 12u) { return; } +out[i] = i32(b0[(((i / 4u) % 3u) * 1u + 1u) * 4u + ((i / 4u) % 1u) * 4u + ((i / 1u) % 4u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fe72b9bbe1c94bc7406073ad754fa678e850e59d --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/121285fc94b4a9f873c60a1d6e0de9d764286607718405fd18288e7ff419c483.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 256u; +if (i >= 256u) { return; } +let x = f32(b0[i]); +if (x * 1.0 > 20.0) { out[i] = f32(x); return; } +out[i] = f32(log(1.0 + exp(x * 1.0)) / 1.0); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1255847146b31732939e573cdb26c483358c1ce07f987bd8c3d3d69ae8c7e141.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1255847146b31732939e573cdb26c483358c1ce07f987bd8c3d3d69ae8c7e141.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..19ce143a6079d1d8518ef9ca512fcc3cdfe1205c --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1255847146b31732939e573cdb26c483358c1ce07f987bd8c3d3d69ae8c7e141.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 24576u; +if (i >= 24576u) { return; } +out[i] = f32(b0[((i / 24576u) % 1u) * 30720u + ((i / 4u) % 6144u) * 5u + (((i / 1u) % 4u) * 1u + 1u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/13237b557466d2fef5bfbc74e7ac77219b9821a53bc0843dbb5d14db8b04d414.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/13237b557466d2fef5bfbc74e7ac77219b9821a53bc0843dbb5d14db8b04d414.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b455658f02601c7cfd2493171e0ae0ce54e635a5 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/13237b557466d2fef5bfbc74e7ac77219b9821a53bc0843dbb5d14db8b04d414.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +out[i] = f32(f32(b0[i]) * f32(b0[i])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4977dd26d91c116b0df884dba3354bba5ff74661 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1326034f3abab37629cd88f869d1b0bafcde724f85c07904b54feaaf71d89c62.wgsl @@ -0,0 +1,44 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var b3: array; +@group(0) @binding(4) var b4: array; +@group(0) @binding(5) var b5: array; +@group(0) @binding(6) var out: array; + +var partial: array; +var delta: f32; +@compute @workgroup_size(128) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let column = group.x; + let head = group.y; + let batch_index = group.z; + let lane = local.x; + var recurrent = 0.0; + if (lane < 128u) { recurrent = f32(b5[((batch_index * 16u + head) * 128u + lane) * 128u + column]); } + for (var token = 0u; token < 4u; token++) { + recurrent *= exp(f32(b3[(batch_index * 4u + token) * 16u + head])); + partial[lane] = 0.0; + if (lane < 128u) { partial[lane] = recurrent * f32(b1[((batch_index * 4u + token) * 16u + head) * 128u + lane]); } + workgroupBarrier(); + for (var stride = 64u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { delta = (f32(b2[((batch_index * 4u + token) * 16u + head) * 128u + column]) - partial[0]) * f32(b4[(batch_index * 4u + token) * 16u + head]); } + workgroupBarrier(); + if (lane < 128u) { recurrent += f32(b1[((batch_index * 4u + token) * 16u + head) * 128u + lane]) * delta; } + partial[lane] = 0.0; + if (lane < 128u) { + partial[lane] = recurrent * f32(b0[((batch_index * 4u + token) * 16u + head) * 128u + lane]) * 0.08838834764831845; + } + workgroupBarrier(); + for (var stride = 64u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { out[((batch_index * 16u + head) * 132u + 128u + token) * 128u + column] = f32(partial[0]); } + workgroupBarrier(); + } + if (lane < 128u) { out[((batch_index * 16u + head) * 132u + lane) * 128u + column] = f32(recurrent); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/13dc73935b9b3b978232043e13babee3877bd53ca572be440de0590dacce8b17.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/13dc73935b9b3b978232043e13babee3877bd53ca572be440de0590dacce8b17.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..de89efb15b0a288331488e51bd02760c3dded87f --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/13dc73935b9b3b978232043e13babee3877bd53ca572be440de0590dacce8b17.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 32u) { return; } +out[i] = f32(b0[0u * 32u + ((i / 32u) % 1u) * 32u + ((i / 32u) % 1u) * 32u + ((i / 1u) % 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/140ebe12d31f1420a8a159b91ae1773390190f496e06e0e5d7b5351f38245068.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/140ebe12d31f1420a8a159b91ae1773390190f496e06e0e5d7b5351f38245068.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b250108bae6829446bb3477c6769fabbd9aff004 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/140ebe12d31f1420a8a159b91ae1773390190f496e06e0e5d7b5351f38245068.wgsl @@ -0,0 +1,96 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var b3: array; +@group(0) @binding(4) var out: array; + +var q_row: array; +var scores: array; +var partial: array; +var limit_shared: u32; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.x; + let head = group.x / 64u; + let row = group.x % 64u; + let kv_head = head / 4u; + for (var d = lane; d < 256u; d += 64u) { q_row[d] = f32(b0[(head * 64u + row) * 256u + d]); } + // Bound the key loop by the last unmasked position so work follows the used context. + var last = 0.0; + for (var j = lane; j < 4096u; j += 64u) { + if (f32(b3[row * 4096u + j]) > -1000.0) { last = max(last, f32(j + 1u)); } + } + partial[lane] = last; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] = max(partial[lane], partial[lane + stride]); } + workgroupBarrier(); + } + if (lane == 0u) { limit_shared = u32(partial[0]); } + let limit = workgroupUniformLoad(&limit_shared); + var running_max = -3.4028234663852886e38; + var running_sum = 0.0; + var acc: array; + for (var start = 0u; start < limit; start += 1024u) { + let count = min(1024u, limit - start); + var tile_max = -3.4028234663852886e38; + for (var t = lane; t < count; t += 64u) { + let j = start + t; + var score = -3.4028234663852886e38; + if (f32(b3[row * 4096u + j]) > -1000.0) { + var dot = 0.0; + for (var d = 0u; d < 256u; d++) { dot += q_row[d] * f32(b1[(((kv_head * 4096u + j) * 256u + d) / 256u) % 4096u * 512u + (((kv_head * 4096u + j) * 256u + d) / 1048576u) % 2u * 256u + (((kv_head * 4096u + j) * 256u + d) / 1u) % 256u * 1u]); } + score = dot * 0.0625 + f32(b3[row * 4096u + j]); + } + scores[t] = score; + tile_max = max(tile_max, score); + } + partial[lane] = tile_max; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] = max(partial[lane], partial[lane + stride]); } + workgroupBarrier(); + } + let next_max = max(running_max, partial[0]); + workgroupBarrier(); + var tile_sum = 0.0; + for (var t = lane; t < count; t += 64u) { + var p = 0.0; + if (scores[t] > -3.0e38) { p = exp(scores[t] - next_max); } + scores[t] = p; + tile_sum += p; + } + partial[lane] = tile_sum; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + var correction = 0.0; + if (running_max > -3.0e38) { correction = exp(running_max - next_max); } + running_sum = running_sum * correction + partial[0]; + running_max = next_max; + for (var c = 0u; c < 4u; c++) { + let d = lane + c * 64u; + var total = acc[c] * correction; + if (d < 256u) { + for (var t = 0u; t < count; t++) { + let p = scores[t]; + if (p != 0.0) { + let j = start + t; + total += p * f32(b2[(((kv_head * 4096u + j) * 256u + d) / 256u) % 4096u * 512u + (((kv_head * 4096u + j) * 256u + d) / 1048576u) % 2u * 256u + (((kv_head * 4096u + j) * 256u + d) / 1u) % 256u * 1u]); + } + } + } + acc[c] = total; + } + workgroupBarrier(); + } + for (var c = 0u; c < 4u; c++) { + let d = lane + c * 64u; + if (d < 256u) { + let i = (head * 64u + row) * 256u + d; + out[i] = f32(acc[c] / running_sum); + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f1322ac39f449a6407142562ed6cad742580c859 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/149c3fdb542967fc038109de3f2b7b2fbee1a6f594baffe28f6f16bafb84d355.wgsl @@ -0,0 +1,24 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var factor: f32; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.x + group.y * 4u; + if (row >= 4u) { return; } + let lane = local.x; + if (lane == 0u) { + var total = 0.0; + for (var j = 0u; j < 1024u; j++) { + let v = f32(b0[row * 1024u + j]); + total += v * v; + } + factor = inverseSqrt(total / 1024.0 + 1e-06); + } + workgroupBarrier(); + for (var p = lane; p < 1024u; p += 64u) { + let i = row * 1024u + p; + out[i] = f32(f32(b0[row * 1024u + p]) * factor * (f32(b1[p]) + 1.0)); + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1545025e0a625ccc73b42ae112c6f9d78a1504c10b759979ae2247984f25f056.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1545025e0a625ccc73b42ae112c6f9d78a1504c10b759979ae2247984f25f056.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..94be53f6bd2811f15de48160517e54acd7fde791 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1545025e0a625ccc73b42ae112c6f9d78a1504c10b759979ae2247984f25f056.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 128u; +if (i >= 128u) { return; } +let coord = (i / 1u) % 32u; +if (coord >= 1u && coord < 34u && (coord - 1u) % 3u == 0u) { out[i] = f32(b0[((i / 128u) % 1u) * 44u + ((i / 32u) % 4u) * 11u + ((((i / 1u) % 32u) - 1u) / 3u) * 1u]); } else { out[i] = f32(b1[i]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/179d5bef469ee5199c2a198580bc5876a1e3d17a7a545c045152d59b246b869f.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/179d5bef469ee5199c2a198580bc5876a1e3d17a7a545c045152d59b246b869f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..383bf5e0bcc9a69ab1a2865b08c47374fdb90045 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/179d5bef469ee5199c2a198580bc5876a1e3d17a7a545c045152d59b246b869f.wgsl @@ -0,0 +1,43 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_0(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b0[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b1[row * 32u + col / 32u]; +} + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let slot = local.x / 16u; + let lane = local.x % 16u; + let i = (group.x + group.y * 1536u) * 4u + slot; + var acc = 0.0; + if (i < 6144u) { + let row = i % 6144u; + for (var w = lane; w < 256u; w += 16u) { + let word = b0[row * 256u + w]; + let first = w * 4u; + var part = 0.0; + for (var e = 0u; e < 4u; e++) { + let col = first + e; + if (col < 1024u) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + part += f32(code) * f32(b2[(i / 6144u) * 1024u + col]); + } + } + acc += part * f32(b1[row * 32u + first / 32u]); + } + } + partial[local.x] = acc; + workgroupBarrier(); + for (var stride = 8u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[local.x] += partial[local.x + stride]; } + workgroupBarrier(); + } + if (lane == 0u && i < 6144u) { out[i] = f32(partial[local.x] + 0.0); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/17ec6bdaf98edf7657b74a25036c2bec806ecbb4280f33fff6afd2e0de5ed377.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/17ec6bdaf98edf7657b74a25036c2bec806ecbb4280f33fff6afd2e0de5ed377.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f8b87c0bd2b9575bbbc8395e5d566f057aeb69dd --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/17ec6bdaf98edf7657b74a25036c2bec806ecbb4280f33fff6afd2e0de5ed377.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +out[i] = f32(f32(b0[i]) * 0.08838834764831843); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/18a0b02a12ebf73203283d267342582d54ba411aeee9cbe21ad89ae890e4cfea.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/18a0b02a12ebf73203283d267342582d54ba411aeee9cbe21ad89ae890e4cfea.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4446cdc4e73df59c1bfe3b5230a9d873261c1cd9 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/18a0b02a12ebf73203283d267342582d54ba411aeee9cbe21ad89ae890e4cfea.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +out[i] = f32(b0[((i / 8192u) % 1u) * 8192u + ((i / 1024u) % 8u) * 256u + ((i / 256u) % 4u) * 2048u + ((i / 1u) % 256u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/18a66ec0875c456c9defd45d15e6ef0c148e64a1618e2825c7f96eadbbe8303a.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/18a66ec0875c456c9defd45d15e6ef0c148e64a1618e2825c7f96eadbbe8303a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..42d192d52f0653490fcfbb1f3fbfc868ec02cdf8 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/18a66ec0875c456c9defd45d15e6ef0c148e64a1618e2825c7f96eadbbe8303a.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 16u) { return; } +out[i] = i32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/18b1cef2bae996f56664d9215180e3bfc77b4e514db52fdf9b6c58be51b53edb.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/18b1cef2bae996f56664d9215180e3bfc77b4e514db52fdf9b6c58be51b53edb.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..460942834da66dc168f11860dfbd38c9cd069806 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/18b1cef2bae996f56664d9215180e3bfc77b4e514db52fdf9b6c58be51b53edb.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 44u) { return; } +out[i] = f32(b0[((i / 44u) % 1u) * 128u + ((i / 11u) % 4u) * 32u + (((i / 1u) % 11u) * 3u + 1u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..41fd5a021bacdb59aa4b8e651a7ca5fc0a5156ce --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/19564c145248f065669be23f5daeeda96e025ee4389bef80983001e83fddb31d.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 64u) { return; } +let x = f32(b0[i]); +if (x * 1.0 > 20.0) { out[i] = f32(x); return; } +out[i] = f32(log(1.0 + exp(x * 1.0)) / 1.0); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1a73b6fceda28f40bba4ae27a986ffa39842db94fa5903ea5bf689a7e854d4f3.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1a73b6fceda28f40bba4ae27a986ffa39842db94fa5903ea5bf689a7e854d4f3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..e2608a38a4a14347c0dd37068bb5040f9280ed0d --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1a73b6fceda28f40bba4ae27a986ffa39842db94fa5903ea5bf689a7e854d4f3.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +let coord = (i / 1u) % 64u; +if (coord >= 0u && coord < 32u) { out[i] = f32(b0[(i / 64u) * 32u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 32u && coord < 64u) { out[i] = f32(b0[(i / 64u) * 32u + (coord - 32u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1ad205080cae29142a3ad87343a0cf3ac4945455398293946110d421ce6d515e.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1ad205080cae29142a3ad87343a0cf3ac4945455398293946110d421ce6d515e.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..04dc360880644481a17005996f0dece2fe491df3 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1ad205080cae29142a3ad87343a0cf3ac4945455398293946110d421ce6d515e.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 128u; +if (i >= 128u) { return; } +out[i] = f32(f32(b0[i]) + (f32(b1[i]) * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1ada8fd10cb41796bc524ec61c2001d2a3ee08e9c062d35d14066088ace72436.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1ada8fd10cb41796bc524ec61c2001d2a3ee08e9c062d35d14066088ace72436.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..dd4007eb615549d9c8da5531f0084b8422606a7b --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1ada8fd10cb41796bc524ec61c2001d2a3ee08e9c062d35d14066088ace72436.wgsl @@ -0,0 +1,43 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_0(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b0[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b1[row * 32u + col / 32u]; +} + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let slot = local.x / 16u; + let lane = local.x % 16u; + let i = (group.x + group.y * 128u) * 4u + slot; + var acc = 0.0; + if (i < 512u) { + let row = i % 512u; + for (var w = lane; w < 256u; w += 16u) { + let word = b0[row * 256u + w]; + let first = w * 4u; + var part = 0.0; + for (var e = 0u; e < 4u; e++) { + let col = first + e; + if (col < 1024u) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + part += f32(code) * f32(b2[(i / 512u) * 1024u + col]); + } + } + acc += part * f32(b1[row * 32u + first / 32u]); + } + } + partial[local.x] = acc; + workgroupBarrier(); + for (var stride = 8u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[local.x] += partial[local.x + stride]; } + workgroupBarrier(); + } + if (lane == 0u && i < 512u) { out[i] = f32(partial[local.x] + 0.0); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1b9f4e330debc7235cfaf8f51d00fec575e09aa6116a45ea001e10ff5f09eb8c.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1b9f4e330debc7235cfaf8f51d00fec575e09aa6116a45ea001e10ff5f09eb8c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..34861471bb74f5ed088ecba3565e1ccc71cad68f --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1b9f4e330debc7235cfaf8f51d00fec575e09aa6116a45ea001e10ff5f09eb8c.wgsl @@ -0,0 +1,17 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1536u; +if (i >= 1536u) { return; } +let batch = i / 512u; +let row = (i / 16u) % 32u; +let col = i % 16u; +var acc = 0.0; +for (var p = 0u; p < 1u; p++) { + acc += f32(b0[(((batch / 1u) % 3u) * 32u) + row * 1u + p]) * f32(b1[(((batch / 1u) % 3u) * 16u) + p * 16u + col]); +} +out[i] = f32(acc); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ee7797b736fd339ff4aefef3b0e224c106a8f7f3 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1c441e95ec3ce8fee9348560af48191844baf402715f575940a0d4eb394c82d2.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 16>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 16u; + var acc: array, 1>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 1u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 4u && col < 1024u) { value = f32(f32(b0[(batch * 4u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 2048u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 1u; r++) { + let a_value = tile_a[local.y * 1u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 1u; r++) { + let out_row = tile_row + local.y * 1u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 4u && out_col < 2048u) { + let i = (batch * 4u + out_row) * 2048u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1d0cbb068fe17123e646934bfa1259f19e5db04e45d72a157d3ea4fa48c81aa0.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1d0cbb068fe17123e646934bfa1259f19e5db04e45d72a157d3ea4fa48c81aa0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0c5bc84965c97bab5f1221cde0dd272918f11497 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1d0cbb068fe17123e646934bfa1259f19e5db04e45d72a157d3ea4fa48c81aa0.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 131072u; +if (i >= 131072u) { return; } +out[i] = f32(b0[((i / 131072u) % 1u) * 131072u + ((i / 16384u) % 8u) * 256u + ((i / 256u) % 64u) * 2048u + ((i / 1u) % 256u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c8d8611c2a923b8f4c053ccadc5d14e813eff70d --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1d7df24388934e162899442d3cc6c6576791aac9f1a5dff86766d840f08fd46e.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 57344u; +if (i >= 57344u) { return; } +let x = f32(b0[i]); let s = f32(f32(x / (1.0 + exp(-x)))); +out[i] = f32(s * f32(b1[i])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ba6aa45f32cb5f338b11181ce9264860b26e3463 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/1ea73621b5acbcc2acfe99cb1cefae97247ee2c199f0b0a8574815ea776ca3bb.wgsl @@ -0,0 +1,21 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 141312u; +if (i >= 141312u) { return; } +let batch = i / 141312u; +let channel = (i / 23u) % 6144u; +let group = channel / 1u; +var acc = 0.0; +for (var c = 0u; c < 1u; c++) { + for (var p = 0u; p < 4u; p++) { + let pos0 = i32((i / 1u) % 23u) * 1 + i32((p / 1u) % 4u) * 1 - 3; + if (pos0 >= 0 && pos0 < 20) { acc += f32(b0[batch * 122880u + (group * 1u + c) * 20u + u32(pos0) * 1u]) * f32(b1[(channel * 1u + c) * 4u + p]); } + } +} +out[i] = f32(acc + 0.0); + +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1597f56a28fb2451f63cf7ddb5821f4054a6a2ac --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/20a4dee9cf34730b4ad77bde7921c189235c4da7d54a3dfb0d589f4d5275c755.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +out[i] = f32(b0[((i / 8192u) % 1u) * 24576u + ((i / 2048u) % 4u) * 6144u + (((i / 1u) % 2048u) * 1u + 4096u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/20fb9e0d0bd2aab34b7f69eac259898e68cac4a0d6b0bda5e95c9126a660336d.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/20fb9e0d0bd2aab34b7f69eac259898e68cac4a0d6b0bda5e95c9126a660336d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5dc4ef605d0aeb8c1e6544aebe9f0ade791f4095 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/20fb9e0d0bd2aab34b7f69eac259898e68cac4a0d6b0bda5e95c9126a660336d.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 64u) { return; } +out[i] = f32(b0[((i / 64u) % 1u) * 128u + ((i / 32u) % 2u) * 64u + ((i / 32u) % 1u) * 64u + (((i / 1u) % 32u) * 1u + 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2182574f59b7d9e38d74b30540c5781827e7da2de1ba6b00327af0b4cb42379d.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2182574f59b7d9e38d74b30540c5781827e7da2de1ba6b00327af0b4cb42379d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a10282b1a7089bcbf3fc3c13bf9435b3d210d5c4 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2182574f59b7d9e38d74b30540c5781827e7da2de1ba6b00327af0b4cb42379d.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 6144u; +if (i >= 6144u) { return; } +out[i] = f32(b0[((i / 6144u) % 1u) * 8192u + ((i / 3072u) % 2u) * 4096u + ((i / 192u) % 16u) * 256u + (((i / 1u) % 192u) * 1u + 64u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/222a2645e3e02d0a4063a63ca1b6fa4c89faee6b76905b20de4dd54ab3077ea3.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/222a2645e3e02d0a4063a63ca1b6fa4c89faee6b76905b20de4dd54ab3077ea3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1447aa7dfa8f3765ba1ecc1e492c6fce42213779 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/222a2645e3e02d0a4063a63ca1b6fa4c89faee6b76905b20de4dd54ab3077ea3.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 3u) { return; } +out[i] = f32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3518ecf411c702ffbcb35673540ae16994c0490c --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/237458dff084b2aa9ccba5578da3b2a0b062951575364d1464250baca3fdac5a.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +out[i] = f32(f32(b0[i]) * f32(b0[i])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8956f47ecd984bb7348dc4845cd50216c449dbec --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/246ea07912fcd8a9073926e863141111eb4091b52cc980a6eef148beed122fc4.wgsl @@ -0,0 +1,24 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var factor: f32; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.x + group.y * 32u; + if (row >= 32u) { return; } + let lane = local.x; + if (lane == 0u) { + var total = 0.0; + for (var j = 0u; j < 256u; j++) { + let v = f32(b0[row * 256u + j]); + total += v * v; + } + factor = inverseSqrt(total / 256.0 + 1e-06); + } + workgroupBarrier(); + for (var p = lane; p < 256u; p += 64u) { + let i = row * 256u + p; + out[i] = f32(f32(b0[row * 256u + p]) * factor * (f32(b1[p]) + 1.0)); + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b060e8fbc7a007a1dbe56dcdbe64e32d2e6dec77 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2606a3e23e0b2e53b86130675899c7dd99329dfab67e9164020c77872dbed134.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 16>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 16u; + var acc: array, 1>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 1u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 16u && col < 1024u) { value = f32(f32(b0[(batch * 16u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 512u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 1u; r++) { + let a_value = tile_a[local.y * 1u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 1u; r++) { + let out_row = tile_row + local.y * 1u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 16u && out_col < 512u) { + let i = (batch * 16u + out_row) * 512u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/26cf6eb3b486d1652247ff130b05ed0e8c8560028eeba59acfc075b976bfdfce.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/26cf6eb3b486d1652247ff130b05ed0e8c8560028eeba59acfc075b976bfdfce.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..7018b9532a9e500c3ab541a52cb08ceeedd4810f --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/26cf6eb3b486d1652247ff130b05ed0e8c8560028eeba59acfc075b976bfdfce.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 30720u; +if (i >= 30720u) { return; } +let coord = (i / 1u) % 5u; +if (coord >= 0u && coord < 4u) { out[i] = f32(b0[(i / 5u) * 4u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 4u && coord < 5u) { out[i] = f32(b1[(i / 5u) * 1u + (coord - 4u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2792ae5841828bd5af938c10096abe385e63bb3ec2a22e850f146739433c12bf.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2792ae5841828bd5af938c10096abe385e63bb3ec2a22e850f146739433c12bf.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0f3bd77e9848ac9445e29bd95a2fdf093837ce6d --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2792ae5841828bd5af938c10096abe385e63bb3ec2a22e850f146739433c12bf.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +out[i] = f32(b0[((i / 8192u) % 1u) * 8192u + ((i / 2048u) % 4u) * 256u + ((i / 256u) % 8u) * 1024u + ((i / 1u) % 256u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/28119aa01122c42bd26ceb37129fe0c95b2a9192163f6dabd90fe058b1e14c54.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/28119aa01122c42bd26ceb37129fe0c95b2a9192163f6dabd90fe058b1e14c54.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..990d32638060f680c62cfdca8e54470607c9e8b0 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/28119aa01122c42bd26ceb37129fe0c95b2a9192163f6dabd90fe058b1e14c54.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 4096u; +if (i >= 4096u) { return; } +let x = f32(b0[i]); +out[i] = f32(sin(x)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..40419b864d89c76a17acf4e27e4a4e644776c533 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/28b69aba1aba878089e1293d59fce422e76487f62cda6ef45c4a1f25cb733f3b.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 65536u; +if (i >= 65536u) { return; } +out[i] = f32(f32(b0[i]) + (f32(b1[i]) * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0f6afccf1386d278d6b3e3a77cb049f5ae6debc3 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2af7517eb26022ada8f8870fd9e1e1ef3da66b9c466336ee18d5295957820b27.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 417792u; +if (i >= 417792u) { return; } +out[i] = f32(b0[((i / 417792u) % 1u) * 436224u + ((i / 68u) % 6144u) * 71u + (((i / 1u) % 68u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2b9e4b05ef5623d721dbe56b5812e551efcc6fd833d81064320e86fe94d8c670.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2b9e4b05ef5623d721dbe56b5812e551efcc6fd833d81064320e86fe94d8c670.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..81712cc83c7e16718c1943957bc24f4abd662317 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2b9e4b05ef5623d721dbe56b5812e551efcc6fd833d81064320e86fe94d8c670.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 4096u; +if (i >= 4096u) { return; } +out[i] = f32(f32(b0[i]) * 1.0); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2bdeb3590b1ba05cf6e9c7e8245f8aa552a12dae84f9578c603c9cee081a82d4.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2bdeb3590b1ba05cf6e9c7e8245f8aa552a12dae84f9578c603c9cee081a82d4.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2b4e4e01f4485a846699c215551f49c2f1452a40 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2bdeb3590b1ba05cf6e9c7e8245f8aa552a12dae84f9578c603c9cee081a82d4.wgsl @@ -0,0 +1,24 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var factor: f32; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.x + group.y * 16u; + if (row >= 16u) { return; } + let lane = local.x; + if (lane == 0u) { + var total = 0.0; + for (var j = 0u; j < 128u; j++) { + let v = f32(b0[row * 128u + j]); + total += v * v; + } + factor = inverseSqrt(total / 128.0 + 1e-06); + } + workgroupBarrier(); + for (var p = lane; p < 128u; p += 64u) { + let i = row * 128u + p; + out[i] = f32(f32(b0[row * 128u + p]) * factor * (f32(b1[p]) + 0.0)); + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2c0cca2a7b0198c317e68992e64f3fe9bba80b67204c2695c20c372471c150d3.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2c0cca2a7b0198c317e68992e64f3fe9bba80b67204c2695c20c372471c150d3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..05c73e3e7e256b462ade35867855db48ad811b19 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2c0cca2a7b0198c317e68992e64f3fe9bba80b67204c2695c20c372471c150d3.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 262144u; +if (i >= 262144u) { return; } +out[i] = f32(f32(b0[((i / 16384u) % 16u) * 128u + ((i / 128u) % 128u) * 1u]) * f32(b1[((i / 16384u) % 16u) * 128u + ((i / 1u) % 128u) * 1u])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2c66dcac8fdd3caa9ff1e2441524070c21c24df9526bee44c49e765a8d4c8222.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2c66dcac8fdd3caa9ff1e2441524070c21c24df9526bee44c49e765a8d4c8222.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..69e4d0a119bf0b802b68acfe45a1cce248ebc3be --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2c66dcac8fdd3caa9ff1e2441524070c21c24df9526bee44c49e765a8d4c8222.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 16384u; +if (i >= 16384u) { return; } +out[i] = f32(b0[((i / 16384u) % 1u) * 32768u + ((i / 2048u) % 8u) * 4096u + ((i / 32u) % 64u) * 64u + (((i / 1u) % 32u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2cd4a533d32097a17641196f0cd315d170c587bc2b1c62b42b79d566d6fbf140.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2cd4a533d32097a17641196f0cd315d170c587bc2b1c62b42b79d566d6fbf140.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..38310e6def55923b47e15db35ea758fb1a108a62 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2cd4a533d32097a17641196f0cd315d170c587bc2b1c62b42b79d566d6fbf140.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 384u; +if (i >= 384u) { return; } +out[i] = f32(b0[((i / 128u) % 3u) * 128u + ((i / 128u) % 1u) * 128u + ((i / 4u) % 32u) * 1u + ((i / 1u) % 4u) * 32u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2e1ce6ef01198780289fd4a35d8f86b5172595ed189cf931e8444091d49e8d3a.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2e1ce6ef01198780289fd4a35d8f86b5172595ed189cf931e8444091d49e8d3a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2269a620c58e84a4903bb1135733ef2385d6af46 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2e1ce6ef01198780289fd4a35d8f86b5172595ed189cf931e8444091d49e8d3a.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 64u) { return; } +out[i] = f32(f32(b0[i]) * 1.0); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2ed0344f526ea256cc4b9261d44a1223b139d0294a07ee072107dc85fa6a4e0f.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2ed0344f526ea256cc4b9261d44a1223b139d0294a07ee072107dc85fa6a4e0f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..dfa75ea46f3ff92d864d51365c324c5f8f06d085 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2ed0344f526ea256cc4b9261d44a1223b139d0294a07ee072107dc85fa6a4e0f.wgsl @@ -0,0 +1,20 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 65536u; +if (i >= 65536u) { return; } +let token = i32(b0[i / 1024u]); +if (token < 131072 || token >= 248320) { out[i] = f32(0.0); return; } +out[i] = f32(dequant_1(u32(token - 131072) * 1024u + i % 1024u)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2eeb54034621ff097879064a60fd4e0dd5d68201e48296ad234413a39579acc8.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2eeb54034621ff097879064a60fd4e0dd5d68201e48296ad234413a39579acc8.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1e46eeebccaa9450e6f35a050b1561b7fc75c63e --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/2eeb54034621ff097879064a60fd4e0dd5d68201e48296ad234413a39579acc8.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +out[i] = f32(b0[((i / 8192u) % 1u) * 16384u + ((i / 2048u) % 4u) * 4096u + ((i / 256u) % 8u) * 512u + (((i / 1u) % 256u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d26f55dc801072951721a041127533286d4ae6fe --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3087f59456d6581e7cdad1a14086af53281e39d2041a6a93387ba6e1daf5f9cb.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 262144u; +if (i >= 262144u) { return; } +out[i] = f32(b0[((i / 262144u) % 1u) * 270336u + ((i / 16384u) % 16u) * 16896u + (((i / 128u) % 128u) * 1u + 0u) * 128u + ((i / 1u) % 128u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3203e08f2c99fa4848f553739b7115a7c677840344f17b1d4391d6130517f5a7.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3203e08f2c99fa4848f553739b7115a7c677840344f17b1d4391d6130517f5a7.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c50d890cb53585ca0b01fc04aaf5f843d4872ab1 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3203e08f2c99fa4848f553739b7115a7c677840344f17b1d4391d6130517f5a7.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 4u) { return; } +out[i] = i32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/327049da924fdfe946e25d5d7e2cd46c6f5d3e801e1e42df106c86077a5354e9.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/327049da924fdfe946e25d5d7e2cd46c6f5d3e801e1e42df106c86077a5354e9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b563d42a464b6b5a7f6e70a0ef367cc96801fc89 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/327049da924fdfe946e25d5d7e2cd46c6f5d3e801e1e42df106c86077a5354e9.wgsl @@ -0,0 +1,20 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 16384u; +if (i >= 16384u) { return; } +let token = i32(b0[i / 1024u]); +if (token < 0 || token >= 131072) { out[i] = f32(0.0); return; } +out[i] = f32(dequant_1(u32(token - 0) * 1024u + i % 1024u)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/32a5027cac5dd969d63fe5d5c88d7c87abb600c43ad1d70e1c56615ea781eff3.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/32a5027cac5dd969d63fe5d5c88d7c87abb600c43ad1d70e1c56615ea781eff3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..acbc0b91943f7202976d3a0119d1ad2f45215e18 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/32a5027cac5dd969d63fe5d5c88d7c87abb600c43ad1d70e1c56615ea781eff3.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 262144u; +if (i >= 262144u) { return; } +out[i] = f32(f32(b0[i]) * f32(b1[((i / 16384u) % 16u) * 128u + ((i / 128u) % 128u) * 1u])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4858cc129b2de7c67aab0a5337e8050b0a3b193d --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/32cb7d055e8ecf035e10d05724e64544393bd3916079992a3d8dce427bb14181.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +var acc = 0.0; +for (var j = 0u; j < 128u; j++) { acc += f32(b0[(((i / 1024u) % 1u) * 131072u + ((i / 16u) % 64u) * 2048u + ((i / 1u) % 16u) * 128u) + (((j / 1u) % 128u) * 1u)]); } +out[i] = f32(acc); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/33e8ec04c66adb4988f562f0f9d3e92d29dfda91a56d0ed62be598c7c737f5ba.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/33e8ec04c66adb4988f562f0f9d3e92d29dfda91a56d0ed62be598c7c737f5ba.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ea57d7db6cc64c5b0d1bd0d43ac21f19e17492ca --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/33e8ec04c66adb4988f562f0f9d3e92d29dfda91a56d0ed62be598c7c737f5ba.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 16384u; +if (i >= 16384u) { return; } +out[i] = f32(b0[((i / 16384u) % 1u) * 32768u + ((i / 2048u) % 8u) * 4096u + ((i / 32u) % 64u) * 64u + (((i / 1u) % 32u) * 1u + 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..50da14799819aa45e6e395deacf4115e0fef2ab4 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3560fe2bb07bf6fb92598b9f9c006029892f16dcb55b418973ed89b2b5086b1b.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 131072u; +if (i >= 131072u) { return; } +out[i] = f32(b0[((i / 131072u) % 1u) * 393216u + ((i / 2048u) % 64u) * 6144u + (((i / 1u) % 2048u) * 1u + 4096u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/359d61f2ba174e54e2a0920fcf61e84e436fd3eaf7e2832e21fdb24094493667.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/359d61f2ba174e54e2a0920fcf61e84e436fd3eaf7e2832e21fdb24094493667.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..385adfb5f3ff92084baf0649aca3a9afbec19314 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/359d61f2ba174e54e2a0920fcf61e84e436fd3eaf7e2832e21fdb24094493667.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 512u; +if (i >= 512u) { return; } +let coord = (i / 1u) % 32u; +if (coord >= 2u && coord < 32u && (coord - 2u) % 3u == 0u) { out[i] = f32(b0[((i / 512u) % 1u) * 160u + ((i / 32u) % 16u) * 10u + ((((i / 1u) % 32u) - 2u) / 3u) * 1u]); } else { out[i] = f32(b1[i]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3622ffa2072335a35d6c9db5d284ed92b3fb5ed304acc2c606408063e7e21dcd.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3622ffa2072335a35d6c9db5d284ed92b3fb5ed304acc2c606408063e7e21dcd.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..024f825c272aa685f0711585490d1ea0e262f385 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3622ffa2072335a35d6c9db5d284ed92b3fb5ed304acc2c606408063e7e21dcd.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +out[i] = f32(b0[((i / 32768u) % 1u) * 65536u + ((i / 2048u) % 16u) * 4096u + ((i / 256u) % 8u) * 512u + (((i / 1u) % 256u) * 1u + 256u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3830061b0fc54c85d3d8a783b927c44484b1a3b68b52b8d8a1b980f7553deebf.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3830061b0fc54c85d3d8a783b927c44484b1a3b68b52b8d8a1b980f7553deebf.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..95c986ed8372eaa483413ddfe7e3a70d877c81b6 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3830061b0fc54c85d3d8a783b927c44484b1a3b68b52b8d8a1b980f7553deebf.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +out[i] = f32(b0[((i / 1024u) % 1u) * 2048u + ((i / 512u) % 2u) * 1024u + ((i / 32u) % 16u) * 64u + (((i / 1u) % 32u) * 1u + 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/38bd527686bcade9ed19a52788ccd711ca668e76854cc6274c480860fce88ff2.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/38bd527686bcade9ed19a52788ccd711ca668e76854cc6274c480860fce88ff2.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..934e1895005499ddd0c97450335aaf4cb8e9454e --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/38bd527686bcade9ed19a52788ccd711ca668e76854cc6274c480860fce88ff2.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +out[i] = f32(b0[((i / 2048u) % 1u) * 6144u + ((i / 2048u) % 1u) * 6144u + (((i / 1u) % 2048u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5e8dc81220dd2f1b86b73186ec9bc17b36194096 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/393ad588d895f6825ccbd78f2e7a2ed539c98ba3c715b93ee7fde02852c9a431.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 16>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 16u; + var acc: array, 1>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 1u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 4u && col < 1024u) { value = f32(f32(b0[(batch * 4u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 16u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 1u; r++) { + let a_value = tile_a[local.y * 1u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 1u; r++) { + let out_row = tile_row + local.y * 1u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 4u && out_col < 16u) { + let i = (batch * 4u + out_row) * 16u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..94055a18d56459161a2b133b2c2bc3d726cebd53 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/39646bdfafd1ab0cb874474e3f1e9d79b57f10582a1804bc1d3762a492fec851.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +out[i] = f32(f32(b0[i]) * f32(b1[((i / 64u) % 64u) * 64u + ((i / 1u) % 64u) * 1u])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..de4fe56b7e7f6725c8af959caf3cf986b1985db0 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/39e41393f5850cc4f0b2e8172a3c53c8c7fb874ceb4df2c6f13f9af6128ad7fc.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +out[i] = f32(f32(b0[i]) * f32(b1[((i / 64u) % 64u) * 64u + ((i / 1u) % 64u) * 1u])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3a76085983a5330bf07a4166ea4975700403f9d8035a8ef450f7f18b9cf47c66.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3a76085983a5330bf07a4166ea4975700403f9d8035a8ef450f7f18b9cf47c66.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5cad6f7d757713f64a9fa63553214201fcfad2bf --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3a76085983a5330bf07a4166ea4975700403f9d8035a8ef450f7f18b9cf47c66.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 131072u; +if (i >= 131072u) { return; } +out[i] = f32(b0[((i / 131072u) % 1u) * 262144u + ((i / 2048u) % 64u) * 4096u + ((i / 256u) % 8u) * 512u + (((i / 1u) % 256u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3b8210d9acf57a3c1cca9e926547520d5994967f27ce92897a15b14137f7da66.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3b8210d9acf57a3c1cca9e926547520d5994967f27ce92897a15b14137f7da66.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3cc6b6a79e115907ddba628a69ca56d1e2fd91f8 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3b8210d9acf57a3c1cca9e926547520d5994967f27ce92897a15b14137f7da66.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 192u; +if (i >= 176u) { return; } +out[i] = f32(b0[((i / 176u) % 1u) * 512u + ((i / 11u) % 16u) * 32u + (((i / 1u) % 11u) * 3u + 1u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3bb850bda95bf77706a2839616a401b773f9de0428d1b427e61fb04e2c35fe74.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3bb850bda95bf77706a2839616a401b773f9de0428d1b427e61fb04e2c35fe74.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a3e680fcc3fb413659faf5613327f7d216420af9 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3bb850bda95bf77706a2839616a401b773f9de0428d1b427e61fb04e2c35fe74.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 128u; +if (i >= 96u) { return; } +let coord = (i / 32u) % 3u; +if (coord == 0u) { out[i] = f32(b0[((i / 32u) % 1u) * 32u + ((i / 32u) % 1u) * 32u + ((i / 1u) % 32u) * 1u]); } else { out[i] = f32(b1[i]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3caf74924884e14997a7fcbb2625a36e3b8a2bcf92da835bf6f787ddae437cdb.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3caf74924884e14997a7fcbb2625a36e3b8a2bcf92da835bf6f787ddae437cdb.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..149c6594ef4e21ce6ec8a070cd41b8da8945b338 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3caf74924884e14997a7fcbb2625a36e3b8a2bcf92da835bf6f787ddae437cdb.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 64u) { return; } +out[i] = i32(b0[((i / 1u) % 16u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3daeb9cba5c07ed47ac55e1c2565aa82f7d8e8409037707e7339d1d96d122144.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3daeb9cba5c07ed47ac55e1c2565aa82f7d8e8409037707e7339d1d96d122144.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5f8915fc662d64438f02eef6c1ea2167f64fff2e --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3daeb9cba5c07ed47ac55e1c2565aa82f7d8e8409037707e7339d1d96d122144.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 512u; +if (i >= 512u) { return; } +let coord = (i / 1u) % 32u; +if (coord >= 1u && coord < 34u && (coord - 1u) % 3u == 0u) { out[i] = f32(b0[((i / 512u) % 1u) * 176u + ((i / 32u) % 16u) * 11u + ((((i / 1u) % 32u) - 1u) / 3u) * 1u]); } else { out[i] = f32(b1[i]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3e34409407100d4858c4c109cf62f3eaf2f4c6081166b0ec8679105e6dfbe7fe.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3e34409407100d4858c4c109cf62f3eaf2f4c6081166b0ec8679105e6dfbe7fe.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8d848be2527138a28929c1abca7db36988b86119 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3e34409407100d4858c4c109cf62f3eaf2f4c6081166b0ec8679105e6dfbe7fe.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 16384u; +if (i >= 16384u) { return; } +let x = f32(b0[i]); +out[i] = f32(-x); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..766ae04c66806d0935847ee89fdcde95fdca3a5b --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3ea275cc11d4d81274b1987190ff6d57c804ff9a5f7fed7dee8c61b19a4c100b.wgsl @@ -0,0 +1,24 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var factor: f32; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.x + group.y * 16u; + if (row >= 16u) { return; } + let lane = local.x; + if (lane == 0u) { + var total = 0.0; + for (var j = 0u; j < 1024u; j++) { + let v = f32(b0[row * 1024u + j]); + total += v * v; + } + factor = inverseSqrt(total / 1024.0 + 1e-06); + } + workgroupBarrier(); + for (var p = lane; p < 1024u; p += 64u) { + let i = row * 1024u + p; + out[i] = f32(f32(b0[row * 1024u + p]) * factor * (f32(b1[p]) + 1.0)); + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3eef2865950af58196f0e6e4eaef47123acfcbd576a636f5278aad5a998e4ba6.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3eef2865950af58196f0e6e4eaef47123acfcbd576a636f5278aad5a998e4ba6.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0fa005f0a9d4752d0d0c7d9ec7209c2e843778fe --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3eef2865950af58196f0e6e4eaef47123acfcbd576a636f5278aad5a998e4ba6.wgsl @@ -0,0 +1,20 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 4096u; +if (i >= 4096u) { return; } +let token = i32(b0[i / 1024u]); +if (token < 0 || token >= 131072) { out[i] = f32(0.0); return; } +out[i] = f32(dequant_1(u32(token - 0) * 1024u + i % 1024u)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3f89d19bd6791657fa702ad00a0517b5f292dbf7d0c196bd5b21c949476d23fd.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3f89d19bd6791657fa702ad00a0517b5f292dbf7d0c196bd5b21c949476d23fd.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..76fef58bda8c3635f8de713e9d2e8807bca9fe33 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3f89d19bd6791657fa702ad00a0517b5f292dbf7d0c196bd5b21c949476d23fd.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 384u; +if (i >= 384u) { return; } +let coord = (i / 128u) % 3u; +if (coord == 0u) { out[i] = f32(b0[((i / 128u) % 1u) * 128u + ((i / 32u) % 4u) * 32u + ((i / 1u) % 32u) * 1u]); } else { out[i] = f32(b1[i]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3fa1efae6d90015ba69cb40cef009f49ad7f7a960018fb92a962707deff4f09e.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3fa1efae6d90015ba69cb40cef009f49ad7f7a960018fb92a962707deff4f09e.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..dd0f6856d5316f3d610290ed98cf1a9c465f554d --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3fa1efae6d90015ba69cb40cef009f49ad7f7a960018fb92a962707deff4f09e.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 10u) { return; } +out[i] = f32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3fe93f80860ec8279a77e1dea55c00d6fe2bd81cb9d0c03624b80e9a999ec449.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3fe93f80860ec8279a77e1dea55c00d6fe2bd81cb9d0c03624b80e9a999ec449.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4229c3cbc4658cfc512abef3db2ffb1dea99f943 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/3fe93f80860ec8279a77e1dea55c00d6fe2bd81cb9d0c03624b80e9a999ec449.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 16>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 16u; + var acc: array, 1>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 1u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 16u && col < 1024u) { value = f32(f32(b0[(batch * 16u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 117248u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 1u; r++) { + let a_value = tile_a[local.y * 1u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 1u; r++) { + let out_row = tile_row + local.y * 1u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 16u && out_col < 117248u) { + let i = (batch * 16u + out_row) * 117248u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a3810cf85ad885e651db3ad41a77356904ea9c05 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/40f115c8647a238bfb3ccb0bb8446bccdc94a0dfadc89111c4920efb2259f41d.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +out[i] = f32(f32(b0[i]) * f32(b1[((i / 64u) % 4u) * 64u + ((i / 1u) % 64u) * 1u])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/42c5af4838daf75fff9d4bb9a5df409511c1d6daed617d05eaf033a977503eac.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/42c5af4838daf75fff9d4bb9a5df409511c1d6daed617d05eaf033a977503eac.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ab4b18691ec8371839272e82e1191ffc00772c73 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/42c5af4838daf75fff9d4bb9a5df409511c1d6daed617d05eaf033a977503eac.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 32u) { return; } +let coord = (i / 1u) % 32u; +if (coord >= 1u && coord < 34u && (coord - 1u) % 3u == 0u) { out[i] = f32(b0[((i / 32u) % 1u) * 11u + ((i / 32u) % 1u) * 11u + ((((i / 1u) % 32u) - 1u) / 3u) * 1u]); } else { out[i] = f32(b1[i]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4406ab1e6e4b9eeb91ace297320aa05a55cc481e02894a4a74b655b4070ade36.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4406ab1e6e4b9eeb91ace297320aa05a55cc481e02894a4a74b655b4070ade36.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3f6f007f45374a356a5ce1118272058bb1a41c01 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4406ab1e6e4b9eeb91ace297320aa05a55cc481e02894a4a74b655b4070ade36.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 98304u; +if (i >= 98304u) { return; } +out[i] = f32(b0[((i / 98304u) % 1u) * 131072u + ((i / 12288u) % 8u) * 16384u + ((i / 192u) % 64u) * 256u + (((i / 1u) % 192u) * 1u + 64u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4540d87eaec90c9eb42badfe0abfa70252f26b6e7de4dd31ee6903fe50e96bd1.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4540d87eaec90c9eb42badfe0abfa70252f26b6e7de4dd31ee6903fe50e96bd1.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a0caa5c08df286ab5dbba084e1b6b90da802badc --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4540d87eaec90c9eb42badfe0abfa70252f26b6e7de4dd31ee6903fe50e96bd1.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 262144u; +if (i >= 262144u) { return; } +out[i] = f32(f32(b0[i]) + (f32(b1[i]) * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/45a1fd3ed514f208718651026bf2ee30dda66b02cdd950dc71e88eb6bdac0b10.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/45a1fd3ed514f208718651026bf2ee30dda66b02cdd950dc71e88eb6bdac0b10.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..695035d8c2fafdd1787dc5d92fff564741c00055 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/45a1fd3ed514f208718651026bf2ee30dda66b02cdd950dc71e88eb6bdac0b10.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +out[i] = f32(f32(b0[i]) * f32(b1[i])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..624963c4b33f008194e17926223b6e3c22aff730 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/466959b3cf0d274924151bdd80808e8451f9d8c2aac1455a44f26fa6d8b44fd0.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 262144u; +if (i >= 262144u) { return; } +out[i] = f32(b0[((i / 262144u) % 1u) * 294912u + ((i / 16384u) % 16u) * 18432u + (((i / 128u) % 128u) * 1u + 0u) * 128u + ((i / 1u) % 128u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/48608b2b7bfbe0b65d9e2d82a94a76683315421acb1bf168b4baf14a3c139f76.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/48608b2b7bfbe0b65d9e2d82a94a76683315421acb1bf168b4baf14a3c139f76.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..e47ed666d2bf202931d9ac16d8cb0dd22e57b6d3 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/48608b2b7bfbe0b65d9e2d82a94a76683315421acb1bf168b4baf14a3c139f76.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 704u; +if (i >= 704u) { return; } +out[i] = f32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..dc3d6070aa02c4dad5f2bf7125c9300ad9860ca8 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4a4eb52af727f11dfbd69b294a8ab3eb5053bd5a2feea3cfdb35ab10bc1a4426.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 64u) { return; } +out[i] = f32(f32(b0[i]) + (1e-06 * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9f6db55c3cd410fd8c73944788ff79bbc5d20ba5 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4b098a370a6c993f2c974bd5afcc00f7e3686a600c458a060c7e7db3193789ba.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 131072u; +if (i >= 131072u) { return; } +out[i] = f32(b0[((i / 131072u) % 1u) * 393216u + ((i / 2048u) % 64u) * 6144u + (((i / 1u) % 2048u) * 1u + 2048u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4b1a7529fd5c57d5d4593d4af58a9a3b545c7d106d18251f6452f8a562b97952.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4b1a7529fd5c57d5d4593d4af58a9a3b545c7d106d18251f6452f8a562b97952.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a2c7dcea92e4a5fa9a904796505efde56c5a3ae5 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4b1a7529fd5c57d5d4593d4af58a9a3b545c7d106d18251f6452f8a562b97952.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 16u) { return; } +out[i] = f32(f32(b0[i]) + (f32(b1[((i / 1u) % 16u) * 1u]) * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4be963f9a4bb928764fd0efe02769e220b0e69defb2f8b22ab28e41a90d3fc14.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4be963f9a4bb928764fd0efe02769e220b0e69defb2f8b22ab28e41a90d3fc14.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..24b926b3b907f9052d46718955950614e3f4d5b2 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4be963f9a4bb928764fd0efe02769e220b0e69defb2f8b22ab28e41a90d3fc14.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 11u) { return; } +out[i] = f32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c56dc651ca20bad5f92c614560567912f07eaab6 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4d05b472f8cb3485606d782585e1061973f1391d129fb765f9cb0f38c5fbb433.wgsl @@ -0,0 +1,44 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var b3: array; +@group(0) @binding(4) var b4: array; +@group(0) @binding(5) var b5: array; +@group(0) @binding(6) var out: array; + +var partial: array; +var delta: f32; +@compute @workgroup_size(128) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let column = group.x; + let head = group.y; + let batch_index = group.z; + let lane = local.x; + var recurrent = 0.0; + if (lane < 128u) { recurrent = f32(b5[((batch_index * 16u + head) * 128u + lane) * 128u + column]); } + for (var token = 0u; token < 16u; token++) { + recurrent *= exp(f32(b3[(batch_index * 16u + token) * 16u + head])); + partial[lane] = 0.0; + if (lane < 128u) { partial[lane] = recurrent * f32(b1[((batch_index * 16u + token) * 16u + head) * 128u + lane]); } + workgroupBarrier(); + for (var stride = 64u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { delta = (f32(b2[((batch_index * 16u + token) * 16u + head) * 128u + column]) - partial[0]) * f32(b4[(batch_index * 16u + token) * 16u + head]); } + workgroupBarrier(); + if (lane < 128u) { recurrent += f32(b1[((batch_index * 16u + token) * 16u + head) * 128u + lane]) * delta; } + partial[lane] = 0.0; + if (lane < 128u) { + partial[lane] = recurrent * f32(b0[((batch_index * 16u + token) * 16u + head) * 128u + lane]) * 0.08838834764831845; + } + workgroupBarrier(); + for (var stride = 64u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { out[((batch_index * 16u + head) * 144u + 128u + token) * 128u + column] = f32(partial[0]); } + workgroupBarrier(); + } + if (lane < 128u) { out[((batch_index * 16u + head) * 144u + lane) * 128u + column] = f32(recurrent); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8a51207a37a0e52b47144cddccc479dbb2363614 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4d46494a55ac0b66fdddd47b9422106ac39e0c7d206f01004e13f93a010e1bf5.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 49152u; +if (i >= 49152u) { return; } +let coord = (i / 1u) % 8u; +if (coord >= 0u && coord < 4u) { out[i] = f32(b0[(i / 8u) * 4u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 4u && coord < 8u) { out[i] = f32(b1[(i / 8u) * 4u + (coord - 4u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4d8ebffbbf193b18c13a5cc17d286bf5c92c5ba53ca022af78a9151ec1e5b9c0.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4d8ebffbbf193b18c13a5cc17d286bf5c92c5ba53ca022af78a9151ec1e5b9c0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..03d07fa428a8b82cd684f25b34ee4bb8835c33e0 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4d8ebffbbf193b18c13a5cc17d286bf5c92c5ba53ca022af78a9151ec1e5b9c0.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +out[i] = f32(b0[((i / 32768u) % 1u) * 131072u + ((i / 4096u) % 8u) * 16384u + ((i / 64u) % 64u) * 256u + (((i / 1u) % 64u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b15719b0d1701c211e699fd46aec102bcf740271 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4ee97e3392700b35c4f59a2b5a3321b23902632cbd6cb2860356f5d97a4352e6.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 64>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 64u; + var acc: array, 4>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 4u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 64u && col < 1024u) { value = f32(f32(b0[(batch * 64u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 16u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 4u; r++) { + let a_value = tile_a[local.y * 4u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 4u; r++) { + let out_row = tile_row + local.y * 4u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 64u && out_col < 16u) { + let i = (batch * 64u + out_row) * 16u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4f2c0d08150d4b97d90c244db2d41baca81484c2630dbbd6d586f2d8a4b07b8b.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4f2c0d08150d4b97d90c244db2d41baca81484c2630dbbd6d586f2d8a4b07b8b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c81e589abef0fba31bf1517192cc0be68ca566c6 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4f2c0d08150d4b97d90c244db2d41baca81484c2630dbbd6d586f2d8a4b07b8b.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 131072u; +if (i >= 131072u) { return; } +out[i] = f32(b0[((i / 131072u) % 1u) * 131072u + ((i / 2048u) % 64u) * 256u + ((i / 256u) % 8u) * 16384u + ((i / 1u) % 256u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4f48f08a8aeb51cc47a10db076a46522807d3b5d35952c9718ce1c1cd760681f.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4f48f08a8aeb51cc47a10db076a46522807d3b5d35952c9718ce1c1cd760681f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1fedbe2d560029fc3bb96a1cacb1cb911d5f0853 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4f48f08a8aeb51cc47a10db076a46522807d3b5d35952c9718ce1c1cd760681f.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 3584u; +if (i >= 3584u) { return; } +let x = f32(b0[i]); let s = f32(f32(x / (1.0 + exp(-x)))); +out[i] = f32(s * f32(b1[i])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4fa24e8f09b07271534341707a7a688af67c7deca1f114bfaeba20a2e066c9f1.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4fa24e8f09b07271534341707a7a688af67c7deca1f114bfaeba20a2e066c9f1.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..83c89e6b3a228b3203eb756839b56f1ff744d6b2 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4fa24e8f09b07271534341707a7a688af67c7deca1f114bfaeba20a2e066c9f1.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1536u; +if (i >= 1536u) { return; } +let coord = (i / 512u) % 3u; +if (coord == 0u) { out[i] = f32(b0[((i / 512u) % 1u) * 512u + ((i / 32u) % 16u) * 32u + ((i / 1u) % 32u) * 1u]); } else { out[i] = f32(b1[i]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4fad29e3e337840e25132f10b3e3d55225f727ae13525ff0b0bdf31e03f09d74.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4fad29e3e337840e25132f10b3e3d55225f727ae13525ff0b0bdf31e03f09d74.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..77c9771680a7b21ef657c6b7303a2db2a2eeb5c4 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/4fad29e3e337840e25132f10b3e3d55225f727ae13525ff0b0bdf31e03f09d74.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 256u; +if (i >= 256u) { return; } +let x = f32(b0[i]); +out[i] = f32(-x); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/50535e4c99d9fb4a272d8260a1415f6b38a7ad1025da64e145f230df697d9b73.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/50535e4c99d9fb4a272d8260a1415f6b38a7ad1025da64e145f230df697d9b73.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9735e1efba0b2297940b84a0830ec5b3b741dbd5 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/50535e4c99d9fb4a272d8260a1415f6b38a7ad1025da64e145f230df697d9b73.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 131072u; +if (i >= 131072u) { return; } +let x = f32(b0[i]); +out[i] = f32(1.0 / (1.0 + exp(-x))); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fc751a3c5de6830de0104027f986f119ec7276f2 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/51fd4d798491c6a80daff4c18b0495926b7f62269dfc3a78a327faa52ba2a759.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 16u) { return; } +let x = f32(b0[i]); +out[i] = f32(exp(x)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..70f8e98145f911a3fcd4871c31d5a135f54d9382 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/52210a661e4d5241f53b3b1caf65456e1988ddd663ded8ead63ee7f7c7e3ff57.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +out[i] = f32(b0[((i / 32768u) % 1u) * 32768u + ((i / 2048u) % 16u) * 128u + ((i / 128u) % 16u) * 2048u + ((i / 1u) % 128u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/52b439507af68e969a6e116d3c8ea5338a4cde5a3fc1e97b729512d17ab2f4f9.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/52b439507af68e969a6e116d3c8ea5338a4cde5a3fc1e97b729512d17ab2f4f9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..67ce4656ea03be5b73e9bffd1ef4a470f3754605 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/52b439507af68e969a6e116d3c8ea5338a4cde5a3fc1e97b729512d17ab2f4f9.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 16u) { return; } +out[i] = i32(b0[((i / 1u) % 4u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3795ac79847c2c36cc2c97b49194d20b16555fe9 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/52d3bb12383ee340fde52b8eead6439f4e4de22278dbdc95434bd8048872f3b8.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 16>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 16u; + var acc: array, 1>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 1u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 4u && col < 1024u) { value = f32(f32(b0[(batch * 4u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 512u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 1u; r++) { + let a_value = tile_a[local.y * 1u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 1u; r++) { + let out_row = tile_row + local.y * 1u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 4u && out_col < 512u) { + let i = (batch * 4u + out_row) * 512u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/55b9c48ba429daf80bbae0b4737f895b475abdc9ddfdc88ab2f894f133ec2bf9.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/55b9c48ba429daf80bbae0b4737f895b475abdc9ddfdc88ab2f894f133ec2bf9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2af2397f376be638d8bdd9858e37047571ac05c7 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/55b9c48ba429daf80bbae0b4737f895b475abdc9ddfdc88ab2f894f133ec2bf9.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 30720u; +if (i >= 30720u) { return; } +let x = f32(b0[i]); +out[i] = f32(x / (1.0 + exp(-x))); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/56167a7e8fb1c67fa72948a8baa71866d54cc775be83f8cfa75a947ac0f58b12.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/56167a7e8fb1c67fa72948a8baa71866d54cc775be83f8cfa75a947ac0f58b12.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..131d933387a2c401768152b9d9a7aebba3045f1d --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/56167a7e8fb1c67fa72948a8baa71866d54cc775be83f8cfa75a947ac0f58b12.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 640u; +if (i >= 640u) { return; } +out[i] = f32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..207ccef3274729b4ec8b34bb1549e283d3949102 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/564b62d37d6845f8049f903d4ba35762c607937d8a9644bfe511fbc7bf90dacd.wgsl @@ -0,0 +1,24 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var factor: f32; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.x + group.y * 64u; + if (row >= 64u) { return; } + let lane = local.x; + if (lane == 0u) { + var total = 0.0; + for (var j = 0u; j < 128u; j++) { + let v = f32(b0[row * 128u + j]); + total += v * v; + } + factor = inverseSqrt(total / 128.0 + 1e-06); + } + workgroupBarrier(); + for (var p = lane; p < 128u; p += 64u) { + let i = row * 128u + p; + out[i] = f32(f32(b0[row * 128u + p]) * factor * (f32(b1[p]) + 0.0)); + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/56d7305021d85f399bfc2a14c0e5a212851de5fba6464ff6078bb170334b7121.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/56d7305021d85f399bfc2a14c0e5a212851de5fba6464ff6078bb170334b7121.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..525f3372ae40b986ed257f1a3968afe95a33988f --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/56d7305021d85f399bfc2a14c0e5a212851de5fba6464ff6078bb170334b7121.wgsl @@ -0,0 +1,43 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_0(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b0[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b1[row * 32u + col / 32u]; +} + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let slot = local.x / 16u; + let lane = local.x % 16u; + let i = (group.x + group.y * 1024u) * 4u + slot; + var acc = 0.0; + if (i < 4096u) { + let row = i % 4096u; + for (var w = lane; w < 256u; w += 16u) { + let word = b0[row * 256u + w]; + let first = w * 4u; + var part = 0.0; + for (var e = 0u; e < 4u; e++) { + let col = first + e; + if (col < 1024u) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + part += f32(code) * f32(b2[(i / 4096u) * 1024u + col]); + } + } + acc += part * f32(b1[row * 32u + first / 32u]); + } + } + partial[local.x] = acc; + workgroupBarrier(); + for (var stride = 8u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[local.x] += partial[local.x + stride]; } + workgroupBarrier(); + } + if (lane == 0u && i < 4096u) { out[i] = f32(partial[local.x] + 0.0); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/58077eb80a5fa253f535c97c82dc07edd48455a17cc52583aeee41c7d1bbd0b3.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/58077eb80a5fa253f535c97c82dc07edd48455a17cc52583aeee41c7d1bbd0b3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..26c3fcd5a0947b3197d818ca3d28084a0e9883c0 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/58077eb80a5fa253f535c97c82dc07edd48455a17cc52583aeee41c7d1bbd0b3.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +out[i] = f32(b0[((i / 1024u) % 1u) * 2048u + ((i / 512u) % 2u) * 1024u + ((i / 32u) % 16u) * 64u + (((i / 1u) % 32u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/59ea88d691b919af03bbb26f0253decfbc35d74f9661cdd3a08570ae85c313d2.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/59ea88d691b919af03bbb26f0253decfbc35d74f9661cdd3a08570ae85c313d2.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5330d07c480a06ace52bbbeb6bf873fea1416d6c --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/59ea88d691b919af03bbb26f0253decfbc35d74f9661cdd3a08570ae85c313d2.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 16u) { return; } +let x = f32(b0[i]); +if (x * 1.0 > 20.0) { out[i] = f32(x); return; } +out[i] = f32(log(1.0 + exp(x * 1.0)) / 1.0); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b4abda1ec110ebd592305d0f5da6c3c12a9cd3b2 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5a3a2ba3e1cb848fbbeeafe66425eae65d9fd196646d15eef291188a8179e044.wgsl @@ -0,0 +1,24 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var factor: f32; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.x + group.y * 256u; + if (row >= 256u) { return; } + let lane = local.x; + if (lane == 0u) { + var total = 0.0; + for (var j = 0u; j < 128u; j++) { + let v = f32(b0[row * 128u + j]); + total += v * v; + } + factor = inverseSqrt(total / 128.0 + 1e-06); + } + workgroupBarrier(); + for (var p = lane; p < 128u; p += 64u) { + let i = row * 128u + p; + out[i] = f32(f32(b0[row * 128u + p]) * factor * (f32(b1[p]) + 0.0)); + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5aa3725287640d289322ab469bd924bd7797f1cef28920ed4526783877c3cf6b.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5aa3725287640d289322ab469bd924bd7797f1cef28920ed4526783877c3cf6b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..6def4ac27ef904a2b09da3fd89c8ca8c0eb2a2db --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5aa3725287640d289322ab469bd924bd7797f1cef28920ed4526783877c3cf6b.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +out[i] = f32(f32(b0[i]) + (f32(b1[i]) * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0bab337250e536ae8fba38e52f533bb044d86428 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5aabb2d12f9b0486434345056b4bd24e4456ed356030141bf18306c0d7524476.wgsl @@ -0,0 +1,17 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +var tile: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(local_invocation_id) local: vec3, @builtin(workgroup_id) group: vec3) { + let batch = group.z; + let input_row = group.y * 16u + local.y; + let input_col = group.x * 16u + local.x; + if (input_row < 6144u && input_col < 64u) { + tile[local.y][local.x] = f32(b0[batch * 393216u + input_row * 64u + input_col]); + } + workgroupBarrier(); + let output_row = group.x * 16u + local.y; + let output_col = group.y * 16u + local.x; + if (output_row < 64u && output_col < 6144u) { out[batch * 393216u + output_row * 6144u + output_col] = f32(tile[local.x][local.y]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5c22e6f60217881c0351e38d19057626a6c2292771c1a83816995245777ac2fd.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5c22e6f60217881c0351e38d19057626a6c2292771c1a83816995245777ac2fd.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a0a200bcf1771e18ac73794fa78eac0151343d48 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5c22e6f60217881c0351e38d19057626a6c2292771c1a83816995245777ac2fd.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 256u; +if (i >= 256u) { return; } +out[i] = f32(b0[((i / 256u) % 1u) * 512u + ((i / 32u) % 8u) * 64u + ((i / 32u) % 1u) * 64u + (((i / 1u) % 32u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5c9a5ae6121b3b10611e09e5f1bc563b70191f2506a2afc0f32de823a9d2dac4.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5c9a5ae6121b3b10611e09e5f1bc563b70191f2506a2afc0f32de823a9d2dac4.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fd1c96b16a3a569b268cd100091060be8f6f800c --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5c9a5ae6121b3b10611e09e5f1bc563b70191f2506a2afc0f32de823a9d2dac4.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 128u; +if (i >= 128u) { return; } +let coord = (i / 1u) % 64u; +if (coord >= 0u && coord < 32u) { out[i] = f32(b0[(i / 64u) * 32u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 32u && coord < 64u) { out[i] = f32(b1[(i / 64u) * 32u + (coord - 32u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..25fb4a60e0d5dc5c38f4f96bec8ea29068f2e356 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5cf28cf1e0aa6d5f0ce563a179880681416b707a70113119e68f1d62dc368c51.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 122880u; +if (i >= 122880u) { return; } +let coord = (i / 1u) % 20u; +if (coord >= 0u && coord < 4u) { out[i] = f32(b0[(i / 20u) * 4u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 4u && coord < 20u) { out[i] = f32(b1[(i / 20u) * 16u + (coord - 4u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5eac5f92aa4d18895369b11adb99ac0d9639d8b615f7240cffe6f6be7046e54a.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5eac5f92aa4d18895369b11adb99ac0d9639d8b615f7240cffe6f6be7046e54a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..44b1efb36c66e24c004b699b692f008a68d352e6 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5eac5f92aa4d18895369b11adb99ac0d9639d8b615f7240cffe6f6be7046e54a.wgsl @@ -0,0 +1,24 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var factor: f32; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.x + group.y * 2u; + if (row >= 2u) { return; } + let lane = local.x; + if (lane == 0u) { + var total = 0.0; + for (var j = 0u; j < 256u; j++) { + let v = f32(b0[row * 256u + j]); + total += v * v; + } + factor = inverseSqrt(total / 256.0 + 1e-06); + } + workgroupBarrier(); + for (var p = lane; p < 256u; p += 64u) { + let i = row * 256u + p; + out[i] = f32(f32(b0[row * 256u + p]) * factor * (f32(b1[p]) + 1.0)); + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5f57b46016eccc3b50e3685e142b7deadce8f074f124e643e6e7015e2cc27357.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5f57b46016eccc3b50e3685e142b7deadce8f074f124e643e6e7015e2cc27357.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..172250df828f8da8336e4e12e54dce7b4361dcca --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5f57b46016eccc3b50e3685e142b7deadce8f074f124e643e6e7015e2cc27357.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 512u; +if (i >= 512u) { return; } +out[i] = f32(b0[0u * 512u + ((i / 512u) % 1u) * 512u + ((i / 32u) % 16u) * 32u + ((i / 1u) % 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5f657afde3ab80a1fc4b6739c0b702bc5bd60e52630762e66b210fbb8b42b52a.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5f657afde3ab80a1fc4b6739c0b702bc5bd60e52630762e66b210fbb8b42b52a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c35d470700b5e1330d30c2ea3f7e0c5d0f77d4a6 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5f657afde3ab80a1fc4b6739c0b702bc5bd60e52630762e66b210fbb8b42b52a.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 384u; +if (i >= 384u) { return; } +out[i] = f32(b0[((i / 384u) % 1u) * 512u + ((i / 192u) % 2u) * 256u + ((i / 192u) % 1u) * 256u + (((i / 1u) % 192u) * 1u + 64u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ceee99ad7864aec44da1701c29c6722ce73360ea --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5f9d7a23bba38f692aefbf2617133c7884ddbe4ea6d985a89a59c1683dcfd6bf.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 256u; +if (i >= 256u) { return; } +out[i] = f32(f32(b0[i]) + (f32(b1[((i / 1u) % 16u) * 1u]) * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5fbfe4534ce5bfff645563250dd536998f537f4d690c4c2fc422723d5af8e3ec.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5fbfe4534ce5bfff645563250dd536998f537f4d690c4c2fc422723d5af8e3ec.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..193a2537b445b31c64b052c52eb4c9d921d70b8a --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5fbfe4534ce5bfff645563250dd536998f537f4d690c4c2fc422723d5af8e3ec.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 192u; +if (i >= 160u) { return; } +out[i] = f32(b0[((i / 160u) % 1u) * 512u + ((i / 10u) % 16u) * 32u + (((i / 1u) % 10u) * 3u + 2u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5fd6e068841f5c054392bef564909dbf086bae9e4030f3625d1143244a08cbbf.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5fd6e068841f5c054392bef564909dbf086bae9e4030f3625d1143244a08cbbf.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d60668fd46afd65337a293687e115ac26feb2dab --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/5fd6e068841f5c054392bef564909dbf086bae9e4030f3625d1143244a08cbbf.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 64u) { return; } +let coord = (i / 1u) % 64u; +if (coord >= 0u && coord < 32u) { out[i] = f32(b0[(i / 64u) * 32u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 32u && coord < 64u) { out[i] = f32(b0[(i / 64u) * 32u + (coord - 32u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b7d98d8c1120f26495f041687f38c922b7d915a0 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/605f51996eef0769d3b456a83f7540262d1761589b55d1fd41fb2b4dd8fa249f.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +out[i] = f32(f32(b0[i]) + (f32(b1[i]) * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/62f30a62c18ed78c9bf37bdc56529accbee2264b9eac931ba0b35a80bb917e7c.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/62f30a62c18ed78c9bf37bdc56529accbee2264b9eac931ba0b35a80bb917e7c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f77c378d9485e77efa187d66dd4704a129bbf949 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/62f30a62c18ed78c9bf37bdc56529accbee2264b9eac931ba0b35a80bb917e7c.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 128u; +if (i >= 128u) { return; } +out[i] = f32(b0[2u * 128u + ((i / 128u) % 1u) * 128u + ((i / 32u) % 4u) * 32u + ((i / 1u) % 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/63396da046b345383e1c7b09aee551d08e4c3c45030ff5c8bb76bbc34bb2abc6.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/63396da046b345383e1c7b09aee551d08e4c3c45030ff5c8bb76bbc34bb2abc6.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a41ec428d6e31c551c8314a9e96300e2fc7194e8 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/63396da046b345383e1c7b09aee551d08e4c3c45030ff5c8bb76bbc34bb2abc6.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +out[i] = f32(b0[((i / 32768u) % 1u) * 32768u + ((i / 4096u) % 8u) * 256u + ((i / 256u) % 16u) * 2048u + ((i / 1u) % 256u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8756d444672e7e7c195431c0a838b00e753a0dce --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6415def2a3f493f401b2577e38c7c41f0adabe06129c5b8058c744b2f93c4c91.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 229376u; +if (i >= 229376u) { return; } +let x = f32(b0[i]); let s = f32(f32(x / (1.0 + exp(-x)))); +out[i] = f32(s * f32(b1[i])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4a5c11bc4a5081edc0742bee29b2392e1756f244 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/64f29e6a6e8772f8cfc7449941fd85f70caff7c7e20f316c39b394500e9a8ce1.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +let x = f32(b0[i]); +out[i] = f32(inverseSqrt(x)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3262b91520b6a20fc3d4af305fa62d20f7f489f3 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/67816e11d408a6230572433502ace2f9a337285eb623360a1cec493e95add139.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +out[i] = f32(b0[((i / 2048u) % 1u) * 2048u + ((i / 1024u) % 2u) * 256u + ((i / 256u) % 4u) * 512u + ((i / 1u) % 256u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/67978e951813ede8016ebf553fb8d3b4ce9e15dd184810544879559208d6b7e1.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/67978e951813ede8016ebf553fb8d3b4ce9e15dd184810544879559208d6b7e1.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3decac1417ec3c8a4798040f6305fb178904c916 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/67978e951813ede8016ebf553fb8d3b4ce9e15dd184810544879559208d6b7e1.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 16u) { return; } +let x = f32(b0[i]); +out[i] = f32(1.0 / (1.0 + exp(-x))); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0b48ebe8b77388bbe25716cf90652b1ca7f6b70f --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/68280baeb493df7bb5708bb14e293795de69aa952e67eb39bf9d9379e5738cb2.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 3584u; + let col = index % 3584u; + let word = b1[row * 448u + col / 8u]; + let code = i32((word >> ((col % 8u) * 4u)) & 15u) - 8; + return f32(code) * b2[row * 112u + col / 32u]; +} + +var tile_a: array, 16>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 16u; + var acc: array, 1>; + for (var k0 = 0u; k0 < 3584u; k0 += 16u) { + for (var e = 0u; e < 1u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 16u && col < 3584u) { value = f32(f32(b0[(batch * 16u + row) * 3584u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 128u) { + let n_local = lane / 2u; + let word_local = lane % 2u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 8u; + var word = 0u; + var scale = 0.0; + if (n_index < 1024u && first < 3584u) { + word = b1[n_index * 448u + first / 8u]; + scale = f32(b2[n_index * 112u + first / 32u]); + } + for (var e = 0u; e < 8u; e++) { + let code = i32((word >> (e * 4u)) & 15u) - 8; + var value = 0.0; + if (first + e < 3584u) { value = f32(code) * scale; } + tile_b[word_local * 8u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 1u; r++) { + let a_value = tile_a[local.y * 1u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 1u; r++) { + let out_row = tile_row + local.y * 1u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 16u && out_col < 1024u) { + let i = (batch * 16u + out_row) * 1024u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/691af757595593839af5e5e97ee656472f6e7c7406f57d1d9faae74816d9fc91.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/691af757595593839af5e5e97ee656472f6e7c7406f57d1d9faae74816d9fc91.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a7942f03138945c79bce1f33264f80e672218280 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/691af757595593839af5e5e97ee656472f6e7c7406f57d1d9faae74816d9fc91.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 6144u; +if (i >= 6144u) { return; } +let coord = (i / 2048u) % 3u; +if (coord == 0u) { out[i] = f32(b0[((i / 2048u) % 1u) * 2048u + ((i / 32u) % 64u) * 32u + ((i / 1u) % 32u) * 1u]); } else { out[i] = f32(b1[i]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6a045508377e47e95d411257071b7ed8f9097fd220c00de6a5167abb548ebe37.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6a045508377e47e95d411257071b7ed8f9097fd220c00de6a5167abb548ebe37.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..6bd34a6b3bb77a79fdcd3a519ca2897dc0b8084d --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6a045508377e47e95d411257071b7ed8f9097fd220c00de6a5167abb548ebe37.wgsl @@ -0,0 +1,21 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 49152u; +if (i >= 49152u) { return; } +let batch = i / 49152u; +let channel = (i / 8u) % 6144u; +let group = channel / 1u; +var acc = 0.0; +for (var c = 0u; c < 1u; c++) { + for (var p = 0u; p < 4u; p++) { + let pos0 = i32((i / 1u) % 8u) * 1 + i32((p / 1u) % 4u) * 1 - 3; + if (pos0 >= 0 && pos0 < 5) { acc += f32(b0[batch * 30720u + (group * 1u + c) * 5u + u32(pos0) * 1u]) * f32(b1[(channel * 1u + c) * 4u + p]); } + } +} +out[i] = f32(acc + 0.0); + +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6a7b2e3778a02c76155d46fa84aeb33f594954e82e09b6b7f2915c0a606eb3f5.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6a7b2e3778a02c76155d46fa84aeb33f594954e82e09b6b7f2915c0a606eb3f5.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..dbb8e11929641cd77e1054eb002c379556e71409 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6a7b2e3778a02c76155d46fa84aeb33f594954e82e09b6b7f2915c0a606eb3f5.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +var acc = 0.0; +for (var j = 0u; j < 128u; j++) { acc += f32(b0[(((i / 2048u) % 1u) * 262144u + ((i / 128u) % 16u) * 16384u + ((i / 1u) % 128u) * 1u) + (((j / 1u) % 128u) * 128u)]); } +out[i] = f32(acc); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6ae4c7f13978e60e3f4ccd5b8172af88d7dc24cfdb236e51b826f13a03d8cd02.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6ae4c7f13978e60e3f4ccd5b8172af88d7dc24cfdb236e51b826f13a03d8cd02.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1d620a8021e4cd0706b810a2dbc2ee727980b6c2 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6ae4c7f13978e60e3f4ccd5b8172af88d7dc24cfdb236e51b826f13a03d8cd02.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 512u; +if (i >= 512u) { return; } +out[i] = f32(b0[((i / 512u) % 1u) * 2048u + ((i / 64u) % 8u) * 256u + ((i / 64u) % 1u) * 256u + (((i / 1u) % 64u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..459016f5ff986a1eb3571496fd063c81574182d8 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6c57b74ac00a3e385675c024ab180b06a05158dcda1371891249af95792f27fc.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +out[i] = f32(b0[((i / 8192u) % 1u) * 8192u + ((i / 4096u) % 2u) * 256u + ((i / 256u) % 16u) * 512u + ((i / 1u) % 256u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6caa35fcab56c639d1859e1b090a7e5bf016dd7f33ecd2373c24ab625de16e09.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6caa35fcab56c639d1859e1b090a7e5bf016dd7f33ecd2373c24ab625de16e09.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5db8d82fad4be98a827f74b0ec3fc41fc2343ddc --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6caa35fcab56c639d1859e1b090a7e5bf016dd7f33ecd2373c24ab625de16e09.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +out[i] = f32(b0[((i / 2048u) % 1u) * 8192u + ((i / 1024u) % 2u) * 4096u + ((i / 64u) % 16u) * 256u + (((i / 1u) % 64u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..aa5ac8b5beae5c1e53d680c578ece9e674631f78 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6d748fa04b83886e523f4d9b826e6865f73b31dfe31385b10bd0b9cc9ec4d1db.wgsl @@ -0,0 +1,13 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { + let i = gid.x + gid.y * 8192u; + if (i >= 8192u) { return; } + let token = (i / 256u) % 16u; + let outer = i / 4096u; + let destination = outer * 1048576u + u32(b1[token]) * 256u + i % 256u; + out[((destination) / 256u) % 4096u * 512u + ((destination) / 1048576u) % 2u * 256u + ((destination) / 1u) % 256u * 1u] = f32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6d82b7ed99a04f03fa0b9d241188d9613fa3ce9ff425cf8ee30a7f4ce6ded13d.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6d82b7ed99a04f03fa0b9d241188d9613fa3ce9ff425cf8ee30a7f4ce6ded13d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4872a52cf4ada5f89c536415428031fbe79b8ce5 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6d82b7ed99a04f03fa0b9d241188d9613fa3ce9ff425cf8ee30a7f4ce6ded13d.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 4096u; +if (i >= 4096u) { return; } +out[i] = f32(b0[((i / 4096u) % 1u) * 8192u + ((i / 2048u) % 2u) * 4096u + ((i / 32u) % 64u) * 64u + (((i / 1u) % 32u) * 1u + 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..92b787efdff3782565df98fa90a1fd6f3fb1c7bf --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/6de6505af6ba3c1c020424d3a99773ff62d2308a2fbf2295ec2ea7e919e4cbb9.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +let x = f32(b0[i]); let s = f32(f32(x / (1.0 + exp(-x)))); +out[i] = f32(s * f32(b1[i])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7002f8fdf09f5124ce2875c3540bce3e77750260556fc1a00a8b13b6c1871c30.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7002f8fdf09f5124ce2875c3540bce3e77750260556fc1a00a8b13b6c1871c30.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5296bf14fb8b70f8480bfd7ccbbb1548cc142bca --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7002f8fdf09f5124ce2875c3540bce3e77750260556fc1a00a8b13b6c1871c30.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 512u; +if (i >= 512u) { return; } +out[i] = f32(f32(b0[i]) * f32(b1[((i / 1u) % 64u) * 1u])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7121b5c223da0b0f28991f55cfed3c3f3390caa0bf65a8e58ae380db281ceb6b.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7121b5c223da0b0f28991f55cfed3c3f3390caa0bf65a8e58ae380db281ceb6b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..69d777ae7caf63f2db83f47bb596dbe1e85a7c4b --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7121b5c223da0b0f28991f55cfed3c3f3390caa0bf65a8e58ae380db281ceb6b.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 11u) { return; } +out[i] = f32(b0[((i / 11u) % 1u) * 32u + ((i / 11u) % 1u) * 32u + (((i / 1u) % 11u) * 3u + 1u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/71595b227aa0f422c12a8e505b326b3ec8b51ab44862b44b4792c54f58b77fc0.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/71595b227aa0f422c12a8e505b326b3ec8b51ab44862b44b4792c54f58b77fc0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..97ba38afedf6b56be3a9a30e8c21271db198c787 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/71595b227aa0f422c12a8e505b326b3ec8b51ab44862b44b4792c54f58b77fc0.wgsl @@ -0,0 +1,43 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_0(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b0[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b1[row * 32u + col / 32u]; +} + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let slot = local.x / 16u; + let lane = local.x % 16u; + let i = (group.x + group.y * 512u) * 4u + slot; + var acc = 0.0; + if (i < 2048u) { + let row = i % 2048u; + for (var w = lane; w < 256u; w += 16u) { + let word = b0[row * 256u + w]; + let first = w * 4u; + var part = 0.0; + for (var e = 0u; e < 4u; e++) { + let col = first + e; + if (col < 1024u) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + part += f32(code) * f32(b2[(i / 2048u) * 1024u + col]); + } + } + acc += part * f32(b1[row * 32u + first / 32u]); + } + } + partial[local.x] = acc; + workgroupBarrier(); + for (var stride = 8u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[local.x] += partial[local.x + stride]; } + workgroupBarrier(); + } + if (lane == 0u && i < 2048u) { out[i] = f32(partial[local.x] + 0.0); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7194f0a413c1d8c6013b45b42b129c0a7557141a31bf3f08946fa1dfd1071393.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7194f0a413c1d8c6013b45b42b129c0a7557141a31bf3f08946fa1dfd1071393.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3fb94c64c32b779b21c8bf8a58afb0556c82b14a --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7194f0a413c1d8c6013b45b42b129c0a7557141a31bf3f08946fa1dfd1071393.wgsl @@ -0,0 +1,24 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +var factor: f32; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let row = group.x + group.y * 1u; + if (row >= 1u) { return; } + let lane = local.x; + if (lane == 0u) { + var total = 0.0; + for (var j = 0u; j < 1024u; j++) { + let v = f32(b0[row * 1024u + j]); + total += v * v; + } + factor = inverseSqrt(total / 1024.0 + 1e-06); + } + workgroupBarrier(); + for (var p = lane; p < 1024u; p += 64u) { + let i = row * 1024u + p; + out[i] = f32(f32(b0[row * 1024u + p]) * factor * (f32(b1[p]) + 1.0)); + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/71d38690bcc7fc30f01f1c8fa2f13a38beff889fcb3ff387861bb70038e92129.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/71d38690bcc7fc30f01f1c8fa2f13a38beff889fcb3ff387861bb70038e92129.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a53c6f1656ae08f4e2c239592dc559814e4bdee4 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/71d38690bcc7fc30f01f1c8fa2f13a38beff889fcb3ff387861bb70038e92129.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 192u; +if (i >= 192u) { return; } +out[i] = i32(b0[(((i / 64u) % 3u) * 1u + 1u) * 64u + ((i / 64u) % 1u) * 64u + ((i / 1u) % 64u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/720dc88131b6bf06e6b09185aa7d5986eb755dcfa60d0e040cbbf621c8f3f761.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/720dc88131b6bf06e6b09185aa7d5986eb755dcfa60d0e040cbbf621c8f3f761.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..77d2c6148b0bb0637e96ebb50a802068507c98e0 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/720dc88131b6bf06e6b09185aa7d5986eb755dcfa60d0e040cbbf621c8f3f761.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 512u; +if (i >= 512u) { return; } +let coord = (i / 1u) % 256u; +if (coord >= 0u && coord < 64u) { out[i] = f32(b0[(i / 256u) * 64u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 64u && coord < 256u) { out[i] = f32(b1[(i / 256u) * 192u + (coord - 64u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5724c8466d198ca39894d20505ee5c1a65748333 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/72474f821f837da4ebaab4cc925fdac78b707346be174a7c033c4c6518935abc.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 131072u; +if (i >= 131072u) { return; } +out[i] = f32(b0[((i / 131072u) % 1u) * 131072u + ((i / 2048u) % 64u) * 128u + ((i / 128u) % 16u) * 8192u + ((i / 1u) % 128u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..6abfb7424e6b3d0131ab97ee69f4553601295d22 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/733b8bc5515900a14d34b3939a10175d25a47b915dd9d016cbe2290ed99ae1f3.wgsl @@ -0,0 +1,17 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +var tile: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(local_invocation_id) local: vec3, @builtin(workgroup_id) group: vec3) { + let batch = group.z; + let input_row = group.y * 16u + local.y; + let input_col = group.x * 16u + local.x; + if (input_row < 64u && input_col < 6144u) { + tile[local.y][local.x] = f32(b0[batch * 393216u + input_row * 6144u + input_col]); + } + workgroupBarrier(); + let output_row = group.x * 16u + local.y; + let output_col = group.y * 16u + local.x; + if (output_row < 6144u && output_col < 64u) { out[batch * 393216u + output_row * 64u + output_col] = f32(tile[local.x][local.y]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0995f2ce93a8810f3254aea52a51e8c4ae81ab52 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/73b861456466f63d096073357cb34eb5f5819861a2961838b6590969f894fa95.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 256u; +if (i >= 256u) { return; } +out[i] = f32(f32(b0[i]) + (1e-06 * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/742db031d819fdf5a756f37e8e824e2dcb2ca3f30efac3ffbd5918181059ae3a.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/742db031d819fdf5a756f37e8e824e2dcb2ca3f30efac3ffbd5918181059ae3a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0879520c73bf4f1395e8a15cc5874fa97dbecde1 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/742db031d819fdf5a756f37e8e824e2dcb2ca3f30efac3ffbd5918181059ae3a.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 262144u; +if (i >= 262144u) { return; } +out[i] = f32(f32(b0[i]) * f32(b1[((i / 16384u) % 16u) * 1u])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/752da00e33abf87b63eba03893a16e7db848f97f1385942d40d8cfb5122d9bd3.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/752da00e33abf87b63eba03893a16e7db848f97f1385942d40d8cfb5122d9bd3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1601c2f1c95920dd739e32da0111d879da60655e --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/752da00e33abf87b63eba03893a16e7db848f97f1385942d40d8cfb5122d9bd3.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 40u) { return; } +out[i] = f32(b0[((i / 40u) % 1u) * 128u + ((i / 10u) % 4u) * 32u + (((i / 1u) % 10u) * 3u + 2u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7559e04808def8b214f2be42df9b8309e743b79b51bb6c0b9260cb3284c92c05.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7559e04808def8b214f2be42df9b8309e743b79b51bb6c0b9260cb3284c92c05.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fe43bd299068ace326f5cef6826b2fb1a448b794 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7559e04808def8b214f2be42df9b8309e743b79b51bb6c0b9260cb3284c92c05.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 64>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 64u; + var acc: array, 4>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 4u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 64u && col < 1024u) { value = f32(f32(b0[(batch * 64u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 117248u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 4u; r++) { + let a_value = tile_a[local.y * 4u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 4u; r++) { + let out_row = tile_row + local.y * 4u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 64u && out_col < 117248u) { + let i = (batch * 64u + out_row) * 117248u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/75984e187047e6f3dd279f8af2f7b4f6360b3afcac502b9324df6f3025cd1fd5.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/75984e187047e6f3dd279f8af2f7b4f6360b3afcac502b9324df6f3025cd1fd5.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3af7cf19899feb1a85911acada7a08a336704aa2 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/75984e187047e6f3dd279f8af2f7b4f6360b3afcac502b9324df6f3025cd1fd5.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 4194240u; +if (i >= 15892480u) { return; } +let coord = (i / 1u) % 248320u; +if (coord >= 0u && coord < 131072u) { out[i] = f32(b0[(i / 248320u) * 131072u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 131072u && coord < 248320u) { out[i] = f32(b1[(i / 248320u) * 117248u + (coord - 131072u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/764403c33cd8994568e068810380cb57e9f2e326dfc09bed63eed1fbed044da1.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/764403c33cd8994568e068810380cb57e9f2e326dfc09bed63eed1fbed044da1.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4d0a0fe1c2ad2ca1647d571a4c09e923d06aee0e --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/764403c33cd8994568e068810380cb57e9f2e326dfc09bed63eed1fbed044da1.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +let x = f32(b0[i]); +out[i] = f32(cos(x)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c36ce90766882622ae62ab165611a02376cc4d8c --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/77f9becbd500abcaf6a8d278cbcb6c7031868ff5bf74f41284940d56894cf9fa.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 64u) { return; } +out[i] = f32(f32(b0[i]) + (f32(b1[((i / 1u) % 16u) * 1u]) * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/781d08cc08a9e49246e543d15b3d95775985fd130f5536c36bac80b33d2b86da.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/781d08cc08a9e49246e543d15b3d95775985fd130f5536c36bac80b33d2b86da.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..1064ad7833d5f65fd417822fb99abd3d47f6c5b9 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/781d08cc08a9e49246e543d15b3d95775985fd130f5536c36bac80b33d2b86da.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 993280u; +if (i >= 993280u) { return; } +let coord = (i / 1u) % 248320u; +if (coord >= 0u && coord < 131072u) { out[i] = f32(b0[(i / 248320u) * 131072u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 131072u && coord < 248320u) { out[i] = f32(b1[(i / 248320u) * 117248u + (coord - 131072u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7928d1b13d79c16f6c0f9d3130cbdb4193cab6efe3a60b38caa1039688fe56de.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7928d1b13d79c16f6c0f9d3130cbdb4193cab6efe3a60b38caa1039688fe56de.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..d569f2c70babe1c72d9fa65e789ae8c5bdd2f048 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7928d1b13d79c16f6c0f9d3130cbdb4193cab6efe3a60b38caa1039688fe56de.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 40u) { return; } +out[i] = f32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..84ae648de9103180e548eb3ee4ab3414443a58b6 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/793acab25c53175e48dbfd1019e7c5f2bd0176d97e2134d7c290a66cfe702c34.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +out[i] = f32(f32(b0[i]) * f32(b1[((i / 2048u) % 16u) * 16u + ((i / 128u) % 16u) * 1u])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..99103af983de95f17fcf5da3f5d3989173f352d3 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/793b4c977feaf93abb2f6e92796d4971ecf3176f10459317fb7a98b028007c01.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +out[i] = f32(f32(b0[i]) + (f32(b1[((i / 1u) % 16u) * 1u]) * 1.0)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7adea2b745b20a2abb96d02a1fca6db5f06f3c5aac850cf8ae1afa6f50868ded.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7adea2b745b20a2abb96d02a1fca6db5f06f3c5aac850cf8ae1afa6f50868ded.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..71689513a7927e5bc0cb73226d7067f7a4cab739 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7adea2b745b20a2abb96d02a1fca6db5f06f3c5aac850cf8ae1afa6f50868ded.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 4096u; +if (i >= 4096u) { return; } +let x = f32(b0[i]); +out[i] = f32(cos(x)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7b7c42c04ea3c6695ddead641e88b82d8f130015e138fc323eec509b5b56cfa9.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7b7c42c04ea3c6695ddead641e88b82d8f130015e138fc323eec509b5b56cfa9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f6badaa4f9e340b199d7d36478143e82014ca9ca --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7b7c42c04ea3c6695ddead641e88b82d8f130015e138fc323eec509b5b56cfa9.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 4096u; +if (i >= 4096u) { return; } +out[i] = f32(b0[((i / 4096u) % 1u) * 8192u + ((i / 2048u) % 2u) * 4096u + ((i / 32u) % 64u) * 64u + (((i / 1u) % 32u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7cdebba634509c1948e33e892c896078e7e8147891d931a0ffcd163eb6f97d46.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7cdebba634509c1948e33e892c896078e7e8147891d931a0ffcd163eb6f97d46.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..26c686a3f1e1cc6cfafcd95b85948b654d6d71ec --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7cdebba634509c1948e33e892c896078e7e8147891d931a0ffcd163eb6f97d46.wgsl @@ -0,0 +1,43 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_0(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b0[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b1[row * 32u + col / 32u]; +} + +var partial: array; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let slot = local.x / 16u; + let lane = local.x % 16u; + let i = (group.x + group.y * 32768u) * 4u + slot; + var acc = 0.0; + if (i < 131072u) { + let row = i % 131072u; + for (var w = lane; w < 256u; w += 16u) { + let word = b0[row * 256u + w]; + let first = w * 4u; + var part = 0.0; + for (var e = 0u; e < 4u; e++) { + let col = first + e; + if (col < 1024u) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + part += f32(code) * f32(b2[(i / 131072u) * 1024u + col]); + } + } + acc += part * f32(b1[row * 32u + first / 32u]); + } + } + partial[local.x] = acc; + workgroupBarrier(); + for (var stride = 8u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[local.x] += partial[local.x + stride]; } + workgroupBarrier(); + } + if (lane == 0u && i < 131072u) { out[i] = f32(partial[local.x] + 0.0); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7dfc17a30d9796026977dc65bedef251d9b2b5fc7eea0ebee019ad956226a513.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7dfc17a30d9796026977dc65bedef251d9b2b5fc7eea0ebee019ad956226a513.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..112dd6f63735fa5f2e14a1fcf7261a224d585ccd --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7dfc17a30d9796026977dc65bedef251d9b2b5fc7eea0ebee019ad956226a513.wgsl @@ -0,0 +1,20 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 65536u; +if (i >= 65536u) { return; } +let token = i32(b0[i / 1024u]); +if (token < 0 || token >= 131072) { out[i] = f32(0.0); return; } +out[i] = f32(dequant_1(u32(token - 0) * 1024u + i % 1024u)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7ea4a12734f1206e9c94a060bf6b24732f46685f0c46a78c3c5fbcd64489d8c3.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7ea4a12734f1206e9c94a060bf6b24732f46685f0c46a78c3c5fbcd64489d8c3.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8e8a838dfb6ac9c617602bd0b2a35a31a9482fbd --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7ea4a12734f1206e9c94a060bf6b24732f46685f0c46a78c3c5fbcd64489d8c3.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1536u; +if (i >= 1536u) { return; } +out[i] = f32(b0[((i / 1536u) % 1u) * 2048u + ((i / 192u) % 8u) * 256u + ((i / 192u) % 1u) * 256u + (((i / 1u) % 192u) * 1u + 64u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2793e22a61404445c507c8e43861e8d94b977c7c --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7f684886a0c031e32c9cf7a4bff880ae692ddc08ddca971ee20d728b4736d0da.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +out[i] = f32(b0[((i / 32768u) % 1u) * 294912u + ((i / 2048u) % 16u) * 18432u + (((i / 128u) % 16u) * 1u + 128u) * 128u + ((i / 1u) % 128u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..97880f9edebfc6372e74ed1a93ed56480649a981 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7f87f0acf51d47032a0dd2dea23e5b1da98f677efa19a65e3b207e7a1b9c7c2a.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 16>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 16u; + var acc: array, 1>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 1u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 4u && col < 1024u) { value = f32(f32(b0[(batch * 4u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 6144u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 1u; r++) { + let a_value = tile_a[local.y * 1u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 1u; r++) { + let out_row = tile_row + local.y * 1u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 4u && out_col < 6144u) { + let i = (batch * 4u + out_row) * 6144u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7fe183583b43b627721ba6853b306288d7ceea4b42b630410f47ea9637220a4b.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7fe183583b43b627721ba6853b306288d7ceea4b42b630410f47ea9637220a4b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..afbf92723660dbb4d364ef0cbeaf40d4cdc4891a --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7fe183583b43b627721ba6853b306288d7ceea4b42b630410f47ea9637220a4b.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +out[i] = f32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..e6d88d0a15c09579597059a1bdcb1e15f863aa03 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/7fed7072c20dc3e66c80b2ac1f737b0e7e70a695e5db34e057f729c945e60ba0.wgsl @@ -0,0 +1,21 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 67584u; +if (i >= 67584u) { return; } +let batch = i / 67584u; +let channel = (i / 11u) % 6144u; +let group = channel / 1u; +var acc = 0.0; +for (var c = 0u; c < 1u; c++) { + for (var p = 0u; p < 4u; p++) { + let pos0 = i32((i / 1u) % 11u) * 1 + i32((p / 1u) % 4u) * 1 - 3; + if (pos0 >= 0 && pos0 < 8) { acc += f32(b0[batch * 49152u + (group * 1u + c) * 8u + u32(pos0) * 1u]) * f32(b1[(channel * 1u + c) * 4u + p]); } + } +} +out[i] = f32(acc + 0.0); + +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..94a7ba7efb5a87b247d6ccad553f5983c0729713 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8044521fb0a89ed8229d56e4981fe85952cec449657de285bc32b4951518ca1a.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 16>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 16u; + var acc: array, 1>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 1u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 16u && col < 1024u) { value = f32(f32(b0[(batch * 16u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 2048u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 1u; r++) { + let a_value = tile_a[local.y * 1u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 1u; r++) { + let out_row = tile_row + local.y * 1u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 16u && out_col < 2048u) { + let i = (batch * 16u + out_row) * 2048u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/80d22263b082805bc9d4499247f92b7a1a783a3e1cdd9c652d304492a1cfd425.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/80d22263b082805bc9d4499247f92b7a1a783a3e1cdd9c652d304492a1cfd425.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9111ceafdc9fd9e44bf84174c85d6fb3627ed1dc --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/80d22263b082805bc9d4499247f92b7a1a783a3e1cdd9c652d304492a1cfd425.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +out[i] = f32(b0[((i / 1024u) % 1u) * 2048u + ((i / 128u) % 8u) * 256u + ((i / 32u) % 4u) * 64u + (((i / 1u) % 32u) * 1u + 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/816861e23c92e156c1ac01f21e4f6dd1cbfc14978b1be8a5994fb45b0f845cb8.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/816861e23c92e156c1ac01f21e4f6dd1cbfc14978b1be8a5994fb45b0f845cb8.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fd8439ddeb258acd3ccd62679898c9ddc4222a6a --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/816861e23c92e156c1ac01f21e4f6dd1cbfc14978b1be8a5994fb45b0f845cb8.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +let x = f32(b0[i]); +out[i] = f32(1.0 / (1.0 + exp(-x))); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8239cfa752a60160251d979de4a5a3e2816d2f6af5094252d1a898e423ad5c46.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8239cfa752a60160251d979de4a5a3e2816d2f6af5094252d1a898e423ad5c46.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..63a88f47a76f97d0a64c790624a628820dc9248f --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8239cfa752a60160251d979de4a5a3e2816d2f6af5094252d1a898e423ad5c46.wgsl @@ -0,0 +1,20 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 16384u; +if (i >= 16384u) { return; } +let token = i32(b0[i / 1024u]); +if (token < 131072 || token >= 248320) { out[i] = f32(0.0); return; } +out[i] = f32(dequant_1(u32(token - 131072) * 1024u + i % 1024u)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8256ea6cdc3500306470c0bb2dcdb44ebdd454dafa218ad6a94dd953b9bbe125.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8256ea6cdc3500306470c0bb2dcdb44ebdd454dafa218ad6a94dd953b9bbe125.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8f492669f518dab2f9593bc47908384f13528421 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8256ea6cdc3500306470c0bb2dcdb44ebdd454dafa218ad6a94dd953b9bbe125.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 4096u; +if (i >= 4096u) { return; } +out[i] = f32(b0[((i / 4096u) % 1u) * 8192u + ((i / 512u) % 8u) * 1024u + ((i / 32u) % 16u) * 64u + (((i / 1u) % 32u) * 1u + 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c3b4bd3bde971e325a9fa9e70305b17851fa3dca --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/827ffce28b59c5790739e7f9624b4ef6ebf876973f7ba942702dd8872f47fcd5.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 417792u; +if (i >= 417792u) { return; } +let x = f32(b0[i]); +out[i] = f32(x / (1.0 + exp(-x))); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f6253e4954ee69ce5aeefbeb74cb19a00eb2ffe1 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/82d700d77c249d637b7592f92c488497797a3961ea85b74ad2e50cccef3ef78f.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +let coord = (i / 1u) % 64u; +if (coord >= 0u && coord < 32u) { out[i] = f32(b0[(i / 64u) * 32u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 32u && coord < 64u) { out[i] = f32(b1[(i / 64u) * 32u + (coord - 32u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/84101fc314579b90d71da4cca3a2c2cdc2074d96d2861c4b8e38c22cc1903138.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/84101fc314579b90d71da4cca3a2c2cdc2074d96d2861c4b8e38c22cc1903138.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..83350b9cd75974291ce80f1a5b1ab6c19754f7ea --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/84101fc314579b90d71da4cca3a2c2cdc2074d96d2861c4b8e38c22cc1903138.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 6144u; +if (i >= 6144u) { return; } +out[i] = f32(b0[((i / 2048u) % 3u) * 2048u + ((i / 2048u) % 1u) * 2048u + ((i / 32u) % 64u) * 1u + ((i / 1u) % 32u) * 64u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/84f24861effe9e2f3b05f98fe6d4e16f4f8204be40a7eeece0668e57060ecc7b.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/84f24861effe9e2f3b05f98fe6d4e16f4f8204be40a7eeece0668e57060ecc7b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b838db5172dc04dbd440a1315507258ee731b118 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/84f24861effe9e2f3b05f98fe6d4e16f4f8204be40a7eeece0668e57060ecc7b.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +out[i] = f32(b0[((i / 32768u) % 1u) * 32768u + ((i / 2048u) % 16u) * 256u + ((i / 256u) % 8u) * 4096u + ((i / 1u) % 256u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..4b8f4a81af2551a5f68d4e7183fb44f770c4ce2b --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/84ff996d084cbc6b8dca21a8f3267cdb1f4c6b9664c24c3002252f05247c4212.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 49152u; +if (i >= 49152u) { return; } +out[i] = f32(b0[((i / 49152u) % 1u) * 67584u + ((i / 8u) % 6144u) * 11u + (((i / 1u) % 8u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/85964f8bbe5a2e4d430930e1050ea76871f0f2453c3cd4ba020002c0562ca680.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/85964f8bbe5a2e4d430930e1050ea76871f0f2453c3cd4ba020002c0562ca680.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8f4401a2be8eee2336c09ba4cc3feb59774b181a --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/85964f8bbe5a2e4d430930e1050ea76871f0f2453c3cd4ba020002c0562ca680.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +let coord = (i / 1u) % 64u; +if (coord >= 0u && coord < 32u) { out[i] = f32(b0[(i / 64u) * 32u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 32u && coord < 64u) { out[i] = f32(b1[(i / 64u) * 32u + (coord - 32u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fb8c0675574869750a3ded79c1434e5eec65cf45 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/863a50c159a82b2128b5eb2b6d5dddfaed1ee5eb643fdd266345be9e0edcaaf9.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 3584u; + let col = index % 3584u; + let word = b1[row * 448u + col / 8u]; + let code = i32((word >> ((col % 8u) * 4u)) & 15u) - 8; + return f32(code) * b2[row * 112u + col / 32u]; +} + +var tile_a: array, 16>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 16u; + var acc: array, 1>; + for (var k0 = 0u; k0 < 3584u; k0 += 16u) { + for (var e = 0u; e < 1u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 4u && col < 3584u) { value = f32(f32(b0[(batch * 4u + row) * 3584u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 128u) { + let n_local = lane / 2u; + let word_local = lane % 2u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 8u; + var word = 0u; + var scale = 0.0; + if (n_index < 1024u && first < 3584u) { + word = b1[n_index * 448u + first / 8u]; + scale = f32(b2[n_index * 112u + first / 32u]); + } + for (var e = 0u; e < 8u; e++) { + let code = i32((word >> (e * 4u)) & 15u) - 8; + var value = 0.0; + if (first + e < 3584u) { value = f32(code) * scale; } + tile_b[word_local * 8u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 1u; r++) { + let a_value = tile_a[local.y * 1u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 1u; r++) { + let out_row = tile_row + local.y * 1u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 4u && out_col < 1024u) { + let i = (batch * 4u + out_row) * 1024u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2addaa5e8e422409a2331ddb072eb71fa919c31f --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/866d9f04c73b553ee9e71b5d4d57644ebb265137bb0c3faada3a72d0807f7065.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 3584u; + let col = index % 3584u; + let word = b1[row * 448u + col / 8u]; + let code = i32((word >> ((col % 8u) * 4u)) & 15u) - 8; + return f32(code) * b2[row * 112u + col / 32u]; +} + +var tile_a: array, 64>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 64u; + var acc: array, 4>; + for (var k0 = 0u; k0 < 3584u; k0 += 16u) { + for (var e = 0u; e < 4u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 64u && col < 3584u) { value = f32(f32(b0[(batch * 64u + row) * 3584u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 128u) { + let n_local = lane / 2u; + let word_local = lane % 2u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 8u; + var word = 0u; + var scale = 0.0; + if (n_index < 1024u && first < 3584u) { + word = b1[n_index * 448u + first / 8u]; + scale = f32(b2[n_index * 112u + first / 32u]); + } + for (var e = 0u; e < 8u; e++) { + let code = i32((word >> (e * 4u)) & 15u) - 8; + var value = 0.0; + if (first + e < 3584u) { value = f32(code) * scale; } + tile_b[word_local * 8u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 4u; r++) { + let a_value = tile_a[local.y * 4u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 4u; r++) { + let out_row = tile_row + local.y * 4u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 64u && out_col < 1024u) { + let i = (batch * 64u + out_row) * 1024u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..39541e77623e367ad5f2dba84b9ea8cf810cb1ee --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/870f571c2825eaa5aa7203a42631ed44e533c50f236974cb4591aa3c73b231a6.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +let x = f32(b0[i]); let s = f32(f32(x / (1.0 + exp(-x)))); +out[i] = f32(s * f32(b1[i])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/896006749045a93d89dca7ed7326aa05c5ff29423400569d34ab2256dea6b0e4.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/896006749045a93d89dca7ed7326aa05c5ff29423400569d34ab2256dea6b0e4.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..153b27098a91332adc33e93c50826711f884ea94 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/896006749045a93d89dca7ed7326aa05c5ff29423400569d34ab2256dea6b0e4.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 128u; +if (i >= 128u) { return; } +out[i] = f32(f32(b0[i]) * f32(b1[((i / 1u) % 64u) * 1u])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/898086c99e03b2cdb78e3e5658c3fff202f1c87d09d8da08659bd1ba85e01b09.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/898086c99e03b2cdb78e3e5658c3fff202f1c87d09d8da08659bd1ba85e01b09.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b5f7727d651290008885c57d7eb3d4d47da40e0d --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/898086c99e03b2cdb78e3e5658c3fff202f1c87d09d8da08659bd1ba85e01b09.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 512u; +if (i >= 512u) { return; } +let coord = (i / 1u) % 64u; +if (coord >= 0u && coord < 32u) { out[i] = f32(b0[(i / 64u) * 32u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 32u && coord < 64u) { out[i] = f32(b1[(i / 64u) * 32u + (coord - 32u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/89f98b79c18e7f4bbbd79cbd8d3e80108d0e1e4407c161d9637b6a88d3fbe06b.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/89f98b79c18e7f4bbbd79cbd8d3e80108d0e1e4407c161d9637b6a88d3fbe06b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..cec2de8e203bef278c3c8e0eb2c7a08106ae1575 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/89f98b79c18e7f4bbbd79cbd8d3e80108d0e1e4407c161d9637b6a88d3fbe06b.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +out[i] = f32(b0[((i / 8192u) % 1u) * 32768u + ((i / 4096u) % 2u) * 16384u + ((i / 64u) % 64u) * 256u + (((i / 1u) % 64u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8a46c4e24d014fd6af88a3f370c4b6161be7c63b6c30211bd62af3dbd87293a8.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8a46c4e24d014fd6af88a3f370c4b6161be7c63b6c30211bd62af3dbd87293a8.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..879d6b86b22a39f635c45604888a324389324c38 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8a46c4e24d014fd6af88a3f370c4b6161be7c63b6c30211bd62af3dbd87293a8.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 256u; +if (i >= 256u) { return; } +out[i] = f32(b0[((i / 256u) % 1u) * 512u + ((i / 32u) % 8u) * 64u + ((i / 32u) % 1u) * 64u + (((i / 1u) % 32u) * 1u + 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8af32f59dfb78c79ac2187ae707f0897b89fb8bb0e3d832e2edac308eefa603c.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8af32f59dfb78c79ac2187ae707f0897b89fb8bb0e3d832e2edac308eefa603c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..393e545d035b3c735b85081828588efa2083e5b8 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8af32f59dfb78c79ac2187ae707f0897b89fb8bb0e3d832e2edac308eefa603c.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1536u; +if (i >= 1536u) { return; } +out[i] = f32(b0[((i / 1536u) % 1u) * 2048u + ((i / 768u) % 2u) * 1024u + ((i / 192u) % 4u) * 256u + (((i / 1u) % 192u) * 1u + 64u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8affc7c0e85127430e406be35ebb5c5354a8dd26ecc4ae9ab6ace80ce6775f46.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8affc7c0e85127430e406be35ebb5c5354a8dd26ecc4ae9ab6ace80ce6775f46.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..f6d4d57d5129f06a73030d35c57119eff88a6513 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8affc7c0e85127430e406be35ebb5c5354a8dd26ecc4ae9ab6ace80ce6775f46.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 131072u; +if (i >= 131072u) { return; } +let coord = (i / 1u) % 256u; +if (coord >= 0u && coord < 64u) { out[i] = f32(b0[(i / 256u) * 64u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 64u && coord < 256u) { out[i] = f32(b1[(i / 256u) * 192u + (coord - 64u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8b540ef0dbe320e8ccc513d0b118e62c6d25dff607a62b8d9de03d1a393a0a5c.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8b540ef0dbe320e8ccc513d0b118e62c6d25dff607a62b8d9de03d1a393a0a5c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..641d8a2382bbd3436044e0d5a0a620319789a160 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8b540ef0dbe320e8ccc513d0b118e62c6d25dff607a62b8d9de03d1a393a0a5c.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 16>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 16u; + var acc: array, 1>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 1u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 16u && col < 1024u) { value = f32(f32(b0[(batch * 16u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 4096u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 1u; r++) { + let a_value = tile_a[local.y * 1u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 1u; r++) { + let out_row = tile_row + local.y * 1u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 16u && out_col < 4096u) { + let i = (batch * 16u + out_row) * 4096u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8b5e4c9b7889f2bccebf60987b6f5b8c23542c3475cd8d78656dc7ecffd81f8d.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8b5e4c9b7889f2bccebf60987b6f5b8c23542c3475cd8d78656dc7ecffd81f8d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..06d3a0be331bebf0bb3a5e2fbcaa41fd764e10ce --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8b5e4c9b7889f2bccebf60987b6f5b8c23542c3475cd8d78656dc7ecffd81f8d.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +out[i] = f32(b0[((i / 2048u) % 1u) * 8192u + ((i / 256u) % 8u) * 1024u + ((i / 64u) % 4u) * 256u + (((i / 1u) % 64u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..de8c2b01add8ed4522e2cbf58e7f08b03967f1ed --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8c2cc45aff0682e734174141c9fcc5edcdac7925ad6128b062befdba21f8695f.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 122880u; +if (i >= 122880u) { return; } +out[i] = f32(b0[((i / 122880u) % 1u) * 141312u + ((i / 20u) % 6144u) * 23u + (((i / 1u) % 20u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8c3a77238461b0ae13318bad504e1e0779f3f47d1782fe0ca78f20112c031146.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8c3a77238461b0ae13318bad504e1e0779f3f47d1782fe0ca78f20112c031146.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..48bca17aba702ff30048b53673eee423547fd053 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8c3a77238461b0ae13318bad504e1e0779f3f47d1782fe0ca78f20112c031146.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 256u; +if (i >= 256u) { return; } +out[i] = f32(b0[((i / 256u) % 1u) * 512u + ((i / 128u) % 2u) * 256u + ((i / 32u) % 4u) * 64u + (((i / 1u) % 32u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3120a16a5ed9e9bba8b6482fa4ae14b95803e24d --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8c92c6f1421c6378452dd1e15eae6d7da76404ac7e9a0b756bace66e6f207d21.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 98304u; +if (i >= 98304u) { return; } +out[i] = f32(b0[((i / 98304u) % 1u) * 122880u + ((i / 16u) % 6144u) * 20u + (((i / 1u) % 16u) * 1u + 4u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8cce7bfa491959f6a7b62ae9cb861d12e2f4ff05227918a4d06bd80669f0a539.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8cce7bfa491959f6a7b62ae9cb861d12e2f4ff05227918a4d06bd80669f0a539.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..afdc69dc467b8f7f632ffe33bddaf90b521e8362 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8cce7bfa491959f6a7b62ae9cb861d12e2f4ff05227918a4d06bd80669f0a539.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 128u; +if (i >= 128u) { return; } +out[i] = f32(b0[1u * 128u + ((i / 128u) % 1u) * 128u + ((i / 32u) % 4u) * 32u + ((i / 1u) % 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..8a01c2daa51364f3125cbebcab9dbf7871446a9a --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8d0a37e497bcb37e2d5a1746fba59fd69680ed8eeada4835c5d0eeef840d57ab.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 16u) { return; } +let x = f32(b0[i]); +out[i] = f32(-x); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3ed04ba7933e584bd12c87751a81db68eb41af8d --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8d2d4993e8058293c7308e58f517619651bbfcfc855acd7e3f5a8cfdfd971dab.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 256u; +if (i >= 256u) { return; } +let x = f32(b0[i]); +out[i] = f32(1.0 / (1.0 + exp(-x))); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8d62df48089497dd19fea48680730210decda42b63f69ecdf9688fc4a43efc58.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8d62df48089497dd19fea48680730210decda42b63f69ecdf9688fc4a43efc58.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..038f09f18992b3725fd5517a3706736493dfe3b5 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8d62df48089497dd19fea48680730210decda42b63f69ecdf9688fc4a43efc58.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 24576u; +if (i >= 24576u) { return; } +out[i] = f32(b0[((i / 24576u) % 1u) * 32768u + ((i / 3072u) % 8u) * 4096u + ((i / 192u) % 16u) * 256u + (((i / 1u) % 192u) * 1u + 64u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..fe8bfc87f8c17b979d4a5ead82a54af189440ac0 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8d924efe3620ed999f0cf4e6adc1df91b7def10b99b5ecb839ceaa662b43c82c.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 256u; +if (i >= 256u) { return; } +var acc = 0.0; +for (var j = 0u; j < 128u; j++) { acc += f32(b0[(((i / 256u) % 1u) * 32768u + ((i / 16u) % 16u) * 2048u + ((i / 1u) % 16u) * 128u) + (((j / 1u) % 128u) * 1u)]); } +out[i] = f32(acc); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8ee286e06c1271162bea74e14b4bd5b1d66890648217c79435e6b3f1d4181ec8.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8ee286e06c1271162bea74e14b4bd5b1d66890648217c79435e6b3f1d4181ec8.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..de988e579ac17256458e6a635d4588a6072950e3 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8ee286e06c1271162bea74e14b4bd5b1d66890648217c79435e6b3f1d4181ec8.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +out[i] = f32(b0[((i / 1024u) % 1u) * 2048u + ((i / 128u) % 8u) * 256u + ((i / 32u) % 4u) * 64u + (((i / 1u) % 32u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3d33151a526384ace0c88643a27cace0eeddec14 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/8f5b5e0b47fdcb6184ffec1c7d8ab07fdda7fad7e2333ca9e5bc36c5f90f85c6.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 2048u; + let col = index % 2048u; + let word = b1[row * 512u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 64u + col / 32u]; +} + +var tile_a: array, 16>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 16u; + var acc: array, 1>; + for (var k0 = 0u; k0 < 2048u; k0 += 16u) { + for (var e = 0u; e < 1u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 16u && col < 2048u) { value = f32(f32(b0[(batch * 16u + row) * 2048u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 1024u && first < 2048u) { + word = b1[n_index * 512u + first / 4u]; + scale = f32(b2[n_index * 64u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 2048u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 1u; r++) { + let a_value = tile_a[local.y * 1u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 1u; r++) { + let out_row = tile_row + local.y * 1u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 16u && out_col < 1024u) { + let i = (batch * 16u + out_row) * 1024u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9044193f1f4724f15ef5fb41040b7fbac5f62fcaea80594737214bf9cd6be41a.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9044193f1f4724f15ef5fb41040b7fbac5f62fcaea80594737214bf9cd6be41a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2e5bc79c78a2b4d66f36a9281f11690ec86715e4 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9044193f1f4724f15ef5fb41040b7fbac5f62fcaea80594737214bf9cd6be41a.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 131072u; +if (i >= 131072u) { return; } +out[i] = f32(b0[((i / 131072u) % 1u) * 262144u + ((i / 2048u) % 64u) * 4096u + ((i / 256u) % 8u) * 512u + (((i / 1u) % 256u) * 1u + 256u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9144d827dae4c52b80abb829178a9ed901ce5dbc6702ad94b0b355e8eaad5ea5.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9144d827dae4c52b80abb829178a9ed901ce5dbc6702ad94b0b355e8eaad5ea5.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0e196cbf9c5f57d582e8784874efe3d9797e27fe --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9144d827dae4c52b80abb829178a9ed901ce5dbc6702ad94b0b355e8eaad5ea5.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 32u) { return; } +let coord = (i / 1u) % 32u; +if (coord >= 2u && coord < 32u && (coord - 2u) % 3u == 0u) { out[i] = f32(b0[((i / 32u) % 1u) * 10u + ((i / 32u) % 1u) * 10u + ((((i / 1u) % 32u) - 2u) / 3u) * 1u]); } else { out[i] = f32(b1[i]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0ac306227a2d2906504dd1b01e28be580c908878 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/928e4759658bc5813ad5b82855d1a8c13e144cdfe50229dfc10b0624d7490858.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +let coord = (i / 1u) % 256u; +if (coord >= 0u && coord < 64u) { out[i] = f32(b0[(i / 256u) * 64u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 64u && coord < 256u) { out[i] = f32(b1[(i / 256u) * 192u + (coord - 64u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c11278cec63a5abd374c47ef747eac9c0192f0bb --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/92d96e9c7fd228d99d89b4db14a742e99e230931d81fa6c3b7fa1caf8d4a1aa4.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 14336u; +if (i >= 14336u) { return; } +let x = f32(b0[i]); let s = f32(f32(x / (1.0 + exp(-x)))); +out[i] = f32(s * f32(b1[i])); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3781d841daf679d470aa9b14d0cf4771d03af034 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/939eb5e3ec5862e0f1d8caec99b814ed6b2bb3a41a132f7149a503e8df582e3c.wgsl @@ -0,0 +1,13 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { + let i = gid.x + gid.y * 2048u; + if (i >= 2048u) { return; } + let token = (i / 256u) % 4u; + let outer = i / 1024u; + let destination = outer * 1048576u + u32(b1[token]) * 256u + i % 256u; + out[((destination) / 256u) % 4096u * 512u + ((destination) / 1048576u) % 2u * 256u + ((destination) / 1u) % 256u * 1u] = f32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/95cba13b595301afbefe20da4a65b93d4d5c5f4a29b9e1d5348d32cb193c6b4b.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/95cba13b595301afbefe20da4a65b93d4d5c5f4a29b9e1d5348d32cb193c6b4b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9c868b10c7c34316c00e41b14857af7bbfe43f60 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/95cba13b595301afbefe20da4a65b93d4d5c5f4a29b9e1d5348d32cb193c6b4b.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 128u; +if (i >= 128u) { return; } +out[i] = f32(b0[0u * 128u + ((i / 128u) % 1u) * 128u + ((i / 32u) % 4u) * 32u + ((i / 1u) % 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9625ca0b071511e5261877fd21627ca354ea0219da86a10887e8054afa15dcfd.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9625ca0b071511e5261877fd21627ca354ea0219da86a10887e8054afa15dcfd.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..e1f0cf834667ef3ff29080e64d8f9d08e6702c3e --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9625ca0b071511e5261877fd21627ca354ea0219da86a10887e8054afa15dcfd.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 32u) { return; } +out[i] = f32(b0[1u * 32u + ((i / 32u) % 1u) * 32u + ((i / 32u) % 1u) * 32u + ((i / 1u) % 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/96924710cde2ae782e03138b4399e2ed2406f8b37c31401f5a79af528c2193fd.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/96924710cde2ae782e03138b4399e2ed2406f8b37c31401f5a79af528c2193fd.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..44ac5689a55d9c197df69bf87f69ab3697c91f87 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/96924710cde2ae782e03138b4399e2ed2406f8b37c31401f5a79af528c2193fd.wgsl @@ -0,0 +1,20 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 4096u; +if (i >= 4096u) { return; } +let token = i32(b0[i / 1024u]); +if (token < 131072 || token >= 248320) { out[i] = f32(0.0); return; } +out[i] = f32(dequant_1(u32(token - 131072) * 1024u + i % 1024u)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..df4a7b88d0870688be525898c6a657b189a070df --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9701df2bcd72ed6a0b72dca81006d9fcc2336e0250fbd66cf37d4fd80667358f.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 131072u; +if (i >= 131072u) { return; } +out[i] = f32(b0[((i / 131072u) % 1u) * 393216u + ((i / 8192u) % 16u) * 24576u + (((i / 128u) % 64u) * 1u + 128u) * 128u + ((i / 1u) % 128u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/970aa8fd65c7e96611a4b6f30cb0b71e24bdaefff0ca117243cab1a5a7e645a9.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/970aa8fd65c7e96611a4b6f30cb0b71e24bdaefff0ca117243cab1a5a7e645a9.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..2f5b9119322f5487af9c8b2b5da3f2cf33a3c6a2 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/970aa8fd65c7e96611a4b6f30cb0b71e24bdaefff0ca117243cab1a5a7e645a9.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 48u) { return; } +out[i] = f32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..002611dac2216301bcefb788edbf31a49a191700 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/97aa9b5cb2a41dbf920a6c15c61964d33dd0d87990a00183ac4b87c614bc376d.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +let coord = (i / 1u) % 256u; +if (coord >= 0u && coord < 64u) { out[i] = f32(b0[(i / 256u) * 64u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 64u && coord < 256u) { out[i] = f32(b1[(i / 256u) * 192u + (coord - 64u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/985bb9ad33654b9adbd4a256d626e346556a10dad514692800da8ebd7910f0f0.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/985bb9ad33654b9adbd4a256d626e346556a10dad514692800da8ebd7910f0f0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..9018c3fdbf086b98282a94876e8a6631eb5442af --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/985bb9ad33654b9adbd4a256d626e346556a10dad514692800da8ebd7910f0f0.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 3u) { return; } +out[i] = i32(b0[(((i / 1u) % 3u) * 1u + 1u) * 1u + ((i / 1u) % 1u) * 1u + ((i / 1u) % 1u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..6f5a0294909820e1048e6c135334639bd4ca22df --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/987d27946334b53b3006cdca1a7158b13a27e01e688047aeafc4b029761b80af.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 24576u; +if (i >= 24576u) { return; } +out[i] = f32(b0[((i / 24576u) % 1u) * 49152u + ((i / 4u) % 6144u) * 8u + (((i / 1u) % 4u) * 1u + 4u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..6f17903f98ea03f7ed94043a81020110ebe8f203 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/99e423f03f9fdfee41f3dcaab73257064ffb2a64110066752208324724505a8b.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 24576u; +if (i >= 24576u) { return; } +out[i] = f32(b0[((i / 24576u) % 1u) * 122880u + ((i / 4u) % 6144u) * 20u + (((i / 1u) % 4u) * 1u + 16u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..29053ed49af749c19ffd9e4c96e2a5cd0e725a17 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9af812cf7d1d31828c41aac6e5c1b9a55ca5cd015f8658d06bb078a5872becfb.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 64>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 64u; + var acc: array, 4>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 4u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 64u && col < 1024u) { value = f32(f32(b0[(batch * 64u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 6144u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 4u; r++) { + let a_value = tile_a[local.y * 4u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 4u; r++) { + let out_row = tile_row + local.y * 4u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 64u && out_col < 6144u) { + let i = (batch * 64u + out_row) * 6144u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9b7697d3e1de37277befbd725307f40a5143faae3dcf43182d4ad89d8572ed42.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9b7697d3e1de37277befbd725307f40a5143faae3dcf43182d4ad89d8572ed42.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ede227a38b1e2608dc62d7a93b0d68ad2fb653fb --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9b7697d3e1de37277befbd725307f40a5143faae3dcf43182d4ad89d8572ed42.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +out[i] = f32(f32(b0[i]) * 1.0); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..e2476989d2a095c8f6a2ba2e76b53b82c3749282 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9cd8236a30e65989f2ed9c7afcfcf2369b6036ade2a7d81c8e203c2789178074.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 417792u; +if (i >= 417792u) { return; } +let coord = (i / 1u) % 68u; +if (coord >= 0u && coord < 4u) { out[i] = f32(b0[(i / 68u) * 4u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 4u && coord < 68u) { out[i] = f32(b1[(i / 68u) * 64u + (coord - 4u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c743049812f0cdac4cac58f95aedc18a7838dbae --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9d12ba7fd89692e99bb291808c5a80ffd5b95478671dde2c936fc3deb8383fa0.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +let x = f32(b0[i]); +if (x * 1.0 > 20.0) { out[i] = f32(x); return; } +out[i] = f32(log(1.0 + exp(x * 1.0)) / 1.0); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9d3d7b777701cd320ca11358e554e1019fb041ee4d765e66071c215242906773.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9d3d7b777701cd320ca11358e554e1019fb041ee4d765e66071c215242906773.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..433c8d99687959f0d2fba22eeee39545cf88e8ca --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9d3d7b777701cd320ca11358e554e1019fb041ee4d765e66071c215242906773.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +out[i] = f32(b0[((i / 32768u) % 1u) * 65536u + ((i / 2048u) % 16u) * 4096u + ((i / 256u) % 8u) * 512u + (((i / 1u) % 256u) * 1u + 0u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..90bdc06efe0122917f683fbb8817bcab407c17a3 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9d504391a987c7be7ccfbb7972c27cf54cd62730c19cbe7949fee2ea059f0c43.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 122880u; +if (i >= 122880u) { return; } +let x = f32(b0[i]); +out[i] = f32(x / (1.0 + exp(-x))); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9d9666074828f92b63a676348da9d7bb80e5a03213f1bc88f4dd8607e8a26711.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9d9666074828f92b63a676348da9d7bb80e5a03213f1bc88f4dd8607e8a26711.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..152165093c579f4cb2a9cab7d17f564b9c725428 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9d9666074828f92b63a676348da9d7bb80e5a03213f1bc88f4dd8607e8a26711.wgsl @@ -0,0 +1,71 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +var tile_a: array, 64>; +var tile_b: array, 16>; +@compute @workgroup_size(16, 16) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.y * 16u + local.x; + let batch = group.z; + let tile_n = group.x * 64u; + let tile_row = group.y * 64u; + var acc: array, 4>; + for (var k0 = 0u; k0 < 1024u; k0 += 16u) { + for (var e = 0u; e < 4u; e++) { + let flat = lane + e * 256u; + let m_local = flat / 16u; + let row = tile_row + m_local; + let col = k0 + flat % 16u; + var value = 0.0; + if (row < 64u && col < 1024u) { value = f32(f32(b0[(batch * 64u + row) * 1024u + col])); } + tile_a[m_local][flat % 16u] = value; + } + if (lane < 256u) { + let n_local = lane / 4u; + let word_local = lane % 4u; + let n_index = tile_n + n_local; + let first = k0 + word_local * 4u; + var word = 0u; + var scale = 0.0; + if (n_index < 131072u && first < 1024u) { + word = b1[n_index * 256u + first / 4u]; + scale = f32(b2[n_index * 32u + first / 32u]); + } + for (var e = 0u; e < 4u; e++) { + let code = i32((word >> (e * 8u)) & 255u) - 128; + var value = 0.0; + if (first + e < 1024u) { value = f32(code) * scale; } + tile_b[word_local * 4u + e][n_local] = value; + } + } + workgroupBarrier(); + for (var kk = 0u; kk < 16u; kk++) { + var b_values: array; + for (var c = 0u; c < 4u; c++) { b_values[c] = tile_b[kk][local.x * 4u + c]; } + for (var r = 0u; r < 4u; r++) { + let a_value = tile_a[local.y * 4u + r][kk]; + for (var c = 0u; c < 4u; c++) { acc[r][c] += a_value * b_values[c]; } + } + } + workgroupBarrier(); + } + for (var r = 0u; r < 4u; r++) { + let out_row = tile_row + local.y * 4u + r; + for (var c = 0u; c < 4u; c++) { + let out_col = tile_n + local.x * 4u + c; + if (out_row < 64u && out_col < 131072u) { + let i = (batch * 64u + out_row) * 131072u + out_col; + out[i] = f32(acc[r][c] + 0.0); + } + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9f040facf0594804dba5b0f10f53b11b7223a1e1705896f95a8f95aa0fff5eec.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9f040facf0594804dba5b0f10f53b11b7223a1e1705896f95a8f95aa0fff5eec.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..31bd75024f015a30070af10901b1b0aa5f0a0a89 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/9f040facf0594804dba5b0f10f53b11b7223a1e1705896f95a8f95aa0fff5eec.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +let coord = (i / 1u) % 32u; +if (coord >= 1u && coord < 34u && (coord - 1u) % 3u == 0u) { out[i] = f32(b0[((i / 2048u) % 1u) * 704u + ((i / 32u) % 64u) * 11u + ((((i / 1u) % 32u) - 1u) / 3u) * 1u]); } else { out[i] = f32(b1[i]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..870c26ddb7eaf8dd8335f56cb6a8270f187dcb59 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a16a539c1a662c846d6c661e572b0b8e19335f9ef41742f784d738bcfd3cd7a6.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 393216u; +if (i >= 393216u) { return; } +out[i] = f32(b0[((i / 393216u) % 1u) * 417792u + ((i / 64u) % 6144u) * 68u + (((i / 1u) % 64u) * 1u + 4u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a1a974f6d8d51fd7c3496ede0966af303ad9f0c0c9ee44cf93ec962d2b0e87e1.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a1a974f6d8d51fd7c3496ede0966af303ad9f0c0c9ee44cf93ec962d2b0e87e1.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..12d890f4a402b0675a5fd2b6d98f661c3f315dd5 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a1a974f6d8d51fd7c3496ede0966af303ad9f0c0c9ee44cf93ec962d2b0e87e1.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 64u) { return; } +let x = f32(b0[i]); +out[i] = f32(-x); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a1cf6b03baf18d389f0a2b95440c17125a8ecc1e633568de985d42cac6e0d1cb.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a1cf6b03baf18d389f0a2b95440c17125a8ecc1e633568de985d42cac6e0d1cb.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..bef42a5c9cf3a75556610ba3eee2eac25cf19401 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a1cf6b03baf18d389f0a2b95440c17125a8ecc1e633568de985d42cac6e0d1cb.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 512u; +if (i >= 512u) { return; } +out[i] = f32(b0[1u * 512u + ((i / 512u) % 1u) * 512u + ((i / 32u) % 16u) * 32u + ((i / 1u) % 32u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..58181f1b0116d2180192b51cc8e05ae53f00587d --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a21c6a35532bc26ccfc46df22f98e1592e3b285e3800544c032edf39425bde1a.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 98304u; +if (i >= 98304u) { return; } +out[i] = f32(b0[((i / 98304u) % 1u) * 98304u + ((i / 16u) % 6144u) * 1u + ((i / 1u) % 16u) * 6144u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..5466706ae8a51fb919a0c6e3bdc851986df7c3b3 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a24673da4fae6d1c49aa143f8425996b2d036db7b0341987dab330c78ead91d7.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 32768u; +if (i >= 32768u) { return; } +out[i] = f32(b0[((i / 32768u) % 1u) * 98304u + ((i / 2048u) % 16u) * 6144u + (((i / 1u) % 2048u) * 1u + 4096u) * 1u]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a278009bfad4d5fb3e44fe3dd3866517a4143ee6da790d52aa62a69eeda0885d.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a278009bfad4d5fb3e44fe3dd3866517a4143ee6da790d52aa62a69eeda0885d.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..ec10cb7b9850a3b53f27ed3d00110e7623b1e874 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a278009bfad4d5fb3e44fe3dd3866517a4143ee6da790d52aa62a69eeda0885d.wgsl @@ -0,0 +1,96 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var b3: array; +@group(0) @binding(4) var out: array; + +var q_row: array; +var scores: array; +var partial: array; +var limit_shared: u32; +@compute @workgroup_size(64) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let lane = local.x; + let head = group.x / 1u; + let row = group.x % 1u; + let kv_head = head / 4u; + for (var d = lane; d < 256u; d += 64u) { q_row[d] = f32(b0[(head * 1u + row) * 256u + d]); } + // Bound the key loop by the last unmasked position so work follows the used context. + var last = 0.0; + for (var j = lane; j < 4096u; j += 64u) { + if (f32(b3[row * 4096u + j]) > -1000.0) { last = max(last, f32(j + 1u)); } + } + partial[lane] = last; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] = max(partial[lane], partial[lane + stride]); } + workgroupBarrier(); + } + if (lane == 0u) { limit_shared = u32(partial[0]); } + let limit = workgroupUniformLoad(&limit_shared); + var running_max = -3.4028234663852886e38; + var running_sum = 0.0; + var acc: array; + for (var start = 0u; start < limit; start += 1024u) { + let count = min(1024u, limit - start); + var tile_max = -3.4028234663852886e38; + for (var t = lane; t < count; t += 64u) { + let j = start + t; + var score = -3.4028234663852886e38; + if (f32(b3[row * 4096u + j]) > -1000.0) { + var dot = 0.0; + for (var d = 0u; d < 256u; d++) { dot += q_row[d] * f32(b1[(((kv_head * 4096u + j) * 256u + d) / 256u) % 4096u * 512u + (((kv_head * 4096u + j) * 256u + d) / 1048576u) % 2u * 256u + (((kv_head * 4096u + j) * 256u + d) / 1u) % 256u * 1u]); } + score = dot * 0.0625 + f32(b3[row * 4096u + j]); + } + scores[t] = score; + tile_max = max(tile_max, score); + } + partial[lane] = tile_max; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] = max(partial[lane], partial[lane + stride]); } + workgroupBarrier(); + } + let next_max = max(running_max, partial[0]); + workgroupBarrier(); + var tile_sum = 0.0; + for (var t = lane; t < count; t += 64u) { + var p = 0.0; + if (scores[t] > -3.0e38) { p = exp(scores[t] - next_max); } + scores[t] = p; + tile_sum += p; + } + partial[lane] = tile_sum; + workgroupBarrier(); + for (var stride = 32u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + var correction = 0.0; + if (running_max > -3.0e38) { correction = exp(running_max - next_max); } + running_sum = running_sum * correction + partial[0]; + running_max = next_max; + for (var c = 0u; c < 4u; c++) { + let d = lane + c * 64u; + var total = acc[c] * correction; + if (d < 256u) { + for (var t = 0u; t < count; t++) { + let p = scores[t]; + if (p != 0.0) { + let j = start + t; + total += p * f32(b2[(((kv_head * 4096u + j) * 256u + d) / 256u) % 4096u * 512u + (((kv_head * 4096u + j) * 256u + d) / 1048576u) % 2u * 256u + (((kv_head * 4096u + j) * 256u + d) / 1u) % 256u * 1u]); + } + } + } + acc[c] = total; + } + workgroupBarrier(); + } + for (var c = 0u; c < 4u; c++) { + let d = lane + c * 64u; + if (d < 256u) { + let i = (head * 1u + row) * 256u + d; + out[i] = f32(acc[c] / running_sum); + } + } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a27baed9f1bab1d547213060cfb98f64b59479a2b7d7e41486ce8d7e7c1888d8.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a27baed9f1bab1d547213060cfb98f64b59479a2b7d7e41486ce8d7e7c1888d8.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..b0cf8eba7008602b137d344fb72876e9637ecb76 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a27baed9f1bab1d547213060cfb98f64b59479a2b7d7e41486ce8d7e7c1888d8.wgsl @@ -0,0 +1,12 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 248320u; +if (i >= 248320u) { return; } +let coord = (i / 1u) % 248320u; +if (coord >= 0u && coord < 131072u) { out[i] = f32(b0[(i / 248320u) * 131072u + (coord - 0u) * 1u + i % 1u]); } +if (coord >= 131072u && coord < 248320u) { out[i] = f32(b1[(i / 248320u) * 117248u + (coord - 131072u) * 1u + i % 1u]); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a2bd65fc2709e093644d6cbf0c292b4d71520009a87bedcf4bca8fcb5f5c15b8.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a2bd65fc2709e093644d6cbf0c292b4d71520009a87bedcf4bca8fcb5f5c15b8.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..a3b7ba18c0aa6c84088bfae203f30b202cf1cedc --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a2bd65fc2709e093644d6cbf0c292b4d71520009a87bedcf4bca8fcb5f5c15b8.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 12u) { return; } +out[i] = f32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..0f992f99ce37f07be233993c8dabc7fe235ba666 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a2e242b97ac8ec546c9f91b5af599c2231915bdabb57d55cdc21f61a49c21007.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +let x = f32(b0[i]); +out[i] = f32(1.0 / (1.0 + exp(-x))); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a37dac611b27158f19aa5bd70c43217286e610e733cbd602b91666b4af42d3ef.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a37dac611b27158f19aa5bd70c43217286e610e733cbd602b91666b4af42d3ef.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..c0eaa7ea6d730c603c8ce3193ea9e00a3ec74433 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a37dac611b27158f19aa5bd70c43217286e610e733cbd602b91666b4af42d3ef.wgsl @@ -0,0 +1,9 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 64u; +if (i >= 1u) { return; } +out[i] = i32(b0[i]); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a3dcb26b08c4e2c0b11e188a362984db75cae2d09ed11e1029d69576b99ed9ac.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a3dcb26b08c4e2c0b11e188a362984db75cae2d09ed11e1029d69576b99ed9ac.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..da84804a4a2ec0c5cef4e946ac5000db8e7e46af --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a3dcb26b08c4e2c0b11e188a362984db75cae2d09ed11e1029d69576b99ed9ac.wgsl @@ -0,0 +1,20 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var out: array; +fn dequant_1(index: u32) -> f32 { + let row = index / 1024u; + let col = index % 1024u; + let word = b1[row * 256u + col / 4u]; + let code = i32((word >> ((col % 4u) * 8u)) & 255u) - 128; + return f32(code) * b2[row * 32u + col / 32u]; +} + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 1024u; +if (i >= 1024u) { return; } +let token = i32(b0[i / 1024u]); +if (token < 131072 || token >= 248320) { out[i] = f32(0.0); return; } +out[i] = f32(dequant_1(u32(token - 131072) * 1024u + i % 1024u)); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a53c1ce51ebaec593ad667efae9603efe3807db8f0ee060a620ac031cc4f2b92.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a53c1ce51ebaec593ad667efae9603efe3807db8f0ee060a620ac031cc4f2b92.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..3ae9e0043808929344cd1e40bf60a2dc532aac06 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a53c1ce51ebaec593ad667efae9603efe3807db8f0ee060a620ac031cc4f2b92.wgsl @@ -0,0 +1,10 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 8192u; +if (i >= 8192u) { return; } +let x = f32(b0[i]); +out[i] = f32(1.0 / (1.0 + exp(-x))); +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..94caaa605cd85dfadbdb72f49366b2f40ac3ddcf --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a62f6b62277fe048d6bac44a25a2e61399e7e2eab103fccd3a23c29a3c827856.wgsl @@ -0,0 +1,44 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var b2: array; +@group(0) @binding(3) var b3: array; +@group(0) @binding(4) var b4: array; +@group(0) @binding(5) var b5: array; +@group(0) @binding(6) var out: array; + +var partial: array; +var delta: f32; +@compute @workgroup_size(128) +fn main(@builtin(workgroup_id) group: vec3, @builtin(local_invocation_id) local: vec3) { + let column = group.x; + let head = group.y; + let batch_index = group.z; + let lane = local.x; + var recurrent = 0.0; + if (lane < 128u) { recurrent = f32(b5[((batch_index * 16u + head) * 128u + lane) * 128u + column]); } + for (var token = 0u; token < 64u; token++) { + recurrent *= exp(f32(b3[(batch_index * 64u + token) * 16u + head])); + partial[lane] = 0.0; + if (lane < 128u) { partial[lane] = recurrent * f32(b1[((batch_index * 64u + token) * 16u + head) * 128u + lane]); } + workgroupBarrier(); + for (var stride = 64u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { delta = (f32(b2[((batch_index * 64u + token) * 16u + head) * 128u + column]) - partial[0]) * f32(b4[(batch_index * 64u + token) * 16u + head]); } + workgroupBarrier(); + if (lane < 128u) { recurrent += f32(b1[((batch_index * 64u + token) * 16u + head) * 128u + lane]) * delta; } + partial[lane] = 0.0; + if (lane < 128u) { + partial[lane] = recurrent * f32(b0[((batch_index * 64u + token) * 16u + head) * 128u + lane]) * 0.08838834764831845; + } + workgroupBarrier(); + for (var stride = 64u; stride > 0u; stride /= 2u) { + if (lane < stride) { partial[lane] += partial[lane + stride]; } + workgroupBarrier(); + } + if (lane == 0u) { out[((batch_index * 16u + head) * 192u + 128u + token) * 128u + column] = f32(partial[0]); } + workgroupBarrier(); + } + if (lane < 128u) { out[((batch_index * 16u + head) * 192u + lane) * 128u + column] = f32(recurrent); } +} diff --git a/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a672c3e158716bde7444c8141f3df1da16da739ccdd2eb2ad60f9d42d0775157.wgsl b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a672c3e158716bde7444c8141f3df1da16da739ccdd2eb2ad60f9d42d0775157.wgsl new file mode 100644 index 0000000000000000000000000000000000000000..aa456f658303b9267eb9faab7425cd740bbdc6f5 --- /dev/null +++ b/qwen35-08b-fp32-int8-attn-int4-mlp-gptq-home-token-major-v2-v5/kernels/a672c3e158716bde7444c8141f3df1da16da739ccdd2eb2ad60f9d42d0775157.wgsl @@ -0,0 +1,11 @@ +@group(0) @binding(0) var b0: array; +@group(0) @binding(1) var b1: array; +@group(0) @binding(2) var out: array; + +@compute @workgroup_size(64) +fn main(@builtin(global_invocation_id) gid: vec3) { +let i = gid.x + gid.y * 2048u; +if (i >= 2048u) { return; } +let coord = (i / 128u) % 1u; +if (coord == 0u) { out[i] = f32(b0[((i / 2048u) % 1u) * 2048u + ((i / 128u) % 16u) * 128u + ((i / 1u) % 128u) * 1u]); } else { out[i] = f32(b1[i]); } +}